| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 1705, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.846153846153847e-07, | |
| "loss": 2.7739, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 7.692307692307694e-07, | |
| "loss": 2.6648, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.153846153846154e-06, | |
| "loss": 2.8447, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.5384615384615387e-06, | |
| "loss": 2.6984, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.9230769230769234e-06, | |
| "loss": 2.7267, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.307692307692308e-06, | |
| "loss": 2.6802, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.6923076923076923e-06, | |
| "loss": 2.5858, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.0769230769230774e-06, | |
| "loss": 2.436, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.4615384615384617e-06, | |
| "loss": 2.3204, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.846153846153847e-06, | |
| "loss": 2.3288, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.230769230769231e-06, | |
| "loss": 2.1069, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.615384615384616e-06, | |
| "loss": 1.9975, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5e-06, | |
| "loss": 2.0434, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.384615384615385e-06, | |
| "loss": 1.9555, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.769230769230769e-06, | |
| "loss": 1.8655, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.153846153846155e-06, | |
| "loss": 1.8597, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.538461538461539e-06, | |
| "loss": 1.809, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.923076923076923e-06, | |
| "loss": 1.7783, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.307692307692308e-06, | |
| "loss": 1.7153, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.692307692307694e-06, | |
| "loss": 1.7193, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.076923076923077e-06, | |
| "loss": 1.6741, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.461538461538462e-06, | |
| "loss": 1.7576, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.846153846153847e-06, | |
| "loss": 1.728, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.230769230769232e-06, | |
| "loss": 1.5899, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.615384615384616e-06, | |
| "loss": 1.6476, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1e-05, | |
| "loss": 1.5837, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.0384615384615386e-05, | |
| "loss": 1.615, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.076923076923077e-05, | |
| "loss": 1.6188, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1153846153846154e-05, | |
| "loss": 1.5616, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1538461538461538e-05, | |
| "loss": 1.6321, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.1923076923076925e-05, | |
| "loss": 1.5472, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.230769230769231e-05, | |
| "loss": 1.4874, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.2692307692307693e-05, | |
| "loss": 1.5321, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3076923076923078e-05, | |
| "loss": 1.5677, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3461538461538463e-05, | |
| "loss": 1.5789, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3846153846153847e-05, | |
| "loss": 1.558, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4230769230769232e-05, | |
| "loss": 1.4183, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4615384615384615e-05, | |
| "loss": 1.5641, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.5000000000000002e-05, | |
| "loss": 1.5382, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.5384615384615387e-05, | |
| "loss": 1.5014, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.576923076923077e-05, | |
| "loss": 1.4522, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.6153846153846154e-05, | |
| "loss": 1.4632, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.653846153846154e-05, | |
| "loss": 1.4419, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.6923076923076924e-05, | |
| "loss": 1.4787, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.730769230769231e-05, | |
| "loss": 1.471, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.7692307692307694e-05, | |
| "loss": 1.4674, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.807692307692308e-05, | |
| "loss": 1.421, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8461538461538465e-05, | |
| "loss": 1.4135, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.8846153846153846e-05, | |
| "loss": 1.5387, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.923076923076923e-05, | |
| "loss": 1.4109, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9615384615384617e-05, | |
| "loss": 1.4647, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2e-05, | |
| "loss": 1.5125, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999981939743593e-05, | |
| "loss": 1.3975, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999927759039606e-05, | |
| "loss": 1.3975, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999837458083738e-05, | |
| "loss": 1.4772, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999711037202162e-05, | |
| "loss": 1.4527, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.9999548496851524e-05, | |
| "loss": 1.4963, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.999934983761892e-05, | |
| "loss": 1.4704, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 1.999911506022192e-05, | |
| "loss": 1.4795, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999884416550855e-05, | |
| "loss": 1.455, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998537154457298e-05, | |
| "loss": 1.4383, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9998194028177105e-05, | |
| "loss": 1.2969, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9997814787907356e-05, | |
| "loss": 1.4535, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9997399435017893e-05, | |
| "loss": 1.4099, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9996947971008986e-05, | |
| "loss": 1.4214, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9996460397511347e-05, | |
| "loss": 1.3191, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999593671628612e-05, | |
| "loss": 1.39, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9995376929224863e-05, | |
| "loss": 1.4771, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999478103834956e-05, | |
| "loss": 1.3907, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.99941490458126e-05, | |
| "loss": 1.4197, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.999348095389677e-05, | |
| "loss": 1.4526, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9992776765015252e-05, | |
| "loss": 1.3396, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9992036481711618e-05, | |
| "loss": 1.4858, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9991260106659798e-05, | |
| "loss": 1.3566, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9990447642664107e-05, | |
| "loss": 1.3758, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9989599092659207e-05, | |
| "loss": 1.4627, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.99887144597101e-05, | |
| "loss": 1.444, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9987793747012125e-05, | |
| "loss": 1.4592, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9986836957890947e-05, | |
| "loss": 1.4336, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9985844095802538e-05, | |
| "loss": 1.3724, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9984815164333163e-05, | |
| "loss": 1.4621, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9983750167199377e-05, | |
| "loss": 1.3457, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998264910824801e-05, | |
| "loss": 1.3869, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.998151199145613e-05, | |
| "loss": 1.4126, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9980338820931074e-05, | |
| "loss": 1.3754, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9979129600910386e-05, | |
| "loss": 1.4018, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9977884335761834e-05, | |
| "loss": 1.3275, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9976603029983383e-05, | |
| "loss": 1.3606, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9975285688203167e-05, | |
| "loss": 1.3201, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9973932315179502e-05, | |
| "loss": 1.3587, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.997254291580083e-05, | |
| "loss": 1.3955, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.997111749508574e-05, | |
| "loss": 1.5024, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 1.9969656058182926e-05, | |
| "loss": 1.4342, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9968158610371164e-05, | |
| "loss": 1.4254, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9966625157059324e-05, | |
| "loss": 1.3801, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9965055703786306e-05, | |
| "loss": 1.4432, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9963450256221066e-05, | |
| "loss": 1.3711, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.996180882016256e-05, | |
| "loss": 1.4053, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9960131401539737e-05, | |
| "loss": 1.3744, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9958418006411522e-05, | |
| "loss": 1.4216, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.995666864096678e-05, | |
| "loss": 1.4584, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9954883311524316e-05, | |
| "loss": 1.4102, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.995306202453283e-05, | |
| "loss": 1.2626, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9951204786570896e-05, | |
| "loss": 1.3363, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9949311604346965e-05, | |
| "loss": 1.3958, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.99473824846993e-05, | |
| "loss": 1.3832, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9945417434595983e-05, | |
| "loss": 1.3707, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9943416461134877e-05, | |
| "loss": 1.4689, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9941379571543597e-05, | |
| "loss": 1.3616, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.9939306773179498e-05, | |
| "loss": 1.3882, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9937198073529626e-05, | |
| "loss": 1.3026, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.993505348021072e-05, | |
| "loss": 1.3696, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.993287300096916e-05, | |
| "loss": 1.2533, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.993065664368094e-05, | |
| "loss": 1.3722, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.992840441635167e-05, | |
| "loss": 1.3511, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.99261163271165e-05, | |
| "loss": 1.4428, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9923792384240134e-05, | |
| "loss": 1.3871, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9921432596116763e-05, | |
| "loss": 1.3628, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.991903697127007e-05, | |
| "loss": 1.3381, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.991660551835317e-05, | |
| "loss": 1.4411, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.99141382461486e-05, | |
| "loss": 1.3643, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9911635163568276e-05, | |
| "loss": 1.3865, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9909096279653455e-05, | |
| "loss": 1.3367, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9906521603574713e-05, | |
| "loss": 1.3274, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9903911144631924e-05, | |
| "loss": 1.3183, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9901264912254186e-05, | |
| "loss": 1.3829, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.9898582915999834e-05, | |
| "loss": 1.3697, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9895865165556375e-05, | |
| "loss": 1.3734, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9893111670740462e-05, | |
| "loss": 1.4211, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9890322441497857e-05, | |
| "loss": 1.4691, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9887497487903406e-05, | |
| "loss": 1.3725, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.988463682016098e-05, | |
| "loss": 1.3903, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.988174044860346e-05, | |
| "loss": 1.3517, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.987880838369269e-05, | |
| "loss": 1.36, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9875840636019435e-05, | |
| "loss": 1.3872, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9872837216303353e-05, | |
| "loss": 1.3582, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9869798135392953e-05, | |
| "loss": 1.3862, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9866723404265546e-05, | |
| "loss": 1.3446, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9863613034027224e-05, | |
| "loss": 1.4092, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.98604670359128e-05, | |
| "loss": 1.4489, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9857285421285785e-05, | |
| "loss": 1.2771, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9854068201638327e-05, | |
| "loss": 1.3381, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9850815388591195e-05, | |
| "loss": 1.3302, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9847526993893712e-05, | |
| "loss": 1.2568, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9844203029423733e-05, | |
| "loss": 1.4585, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.984084350718758e-05, | |
| "loss": 1.4134, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9837448439320027e-05, | |
| "loss": 1.3563, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9834017838084234e-05, | |
| "loss": 1.3011, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9830551715871703e-05, | |
| "loss": 1.396, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.982705008520225e-05, | |
| "loss": 1.2511, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9823512958723942e-05, | |
| "loss": 1.4065, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.981994034921306e-05, | |
| "loss": 1.3925, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9816332269574056e-05, | |
| "loss": 1.3561, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9812688732839497e-05, | |
| "loss": 1.3132, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9809009752170022e-05, | |
| "loss": 1.3534, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9805295340854303e-05, | |
| "loss": 1.4454, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.9801545512308982e-05, | |
| "loss": 1.3429, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.979776028007863e-05, | |
| "loss": 1.4072, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.97939396578357e-05, | |
| "loss": 1.28, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.979008365938048e-05, | |
| "loss": 1.359, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.978619229864103e-05, | |
| "loss": 1.2664, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9782265589673146e-05, | |
| "loss": 1.3265, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9778303546660304e-05, | |
| "loss": 1.3663, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9774306183913602e-05, | |
| "loss": 1.3191, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9770273515871724e-05, | |
| "loss": 1.4161, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.976620555710087e-05, | |
| "loss": 1.3269, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9762102322294716e-05, | |
| "loss": 1.4626, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9757963826274357e-05, | |
| "loss": 1.3634, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9753790083988256e-05, | |
| "loss": 1.3309, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.974958111051218e-05, | |
| "loss": 1.276, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.974533692104916e-05, | |
| "loss": 1.3402, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9741057530929425e-05, | |
| "loss": 1.3347, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.973674295561035e-05, | |
| "loss": 1.3164, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9732393210676405e-05, | |
| "loss": 1.4294, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.972800831183909e-05, | |
| "loss": 1.2978, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9723588274936882e-05, | |
| "loss": 1.3527, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9719133115935188e-05, | |
| "loss": 1.3472, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.9714642850926264e-05, | |
| "loss": 1.2505, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.971011749612918e-05, | |
| "loss": 1.3672, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.970555706788975e-05, | |
| "loss": 1.3688, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9700961582680476e-05, | |
| "loss": 1.3013, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9696331057100484e-05, | |
| "loss": 1.3348, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9691665507875467e-05, | |
| "loss": 1.3899, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9686964951857636e-05, | |
| "loss": 1.3406, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9682229406025635e-05, | |
| "loss": 1.2191, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.96774588874845e-05, | |
| "loss": 1.3919, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9672653413465584e-05, | |
| "loss": 1.4293, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9667813001326512e-05, | |
| "loss": 1.3165, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9662937668551097e-05, | |
| "loss": 1.3989, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9658027432749293e-05, | |
| "loss": 1.2933, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9653082311657124e-05, | |
| "loss": 1.2817, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.964810232313662e-05, | |
| "loss": 1.3737, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9643087485175752e-05, | |
| "loss": 1.2853, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.9638037815888378e-05, | |
| "loss": 1.3151, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.963295333351416e-05, | |
| "loss": 1.3505, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.96278340564185e-05, | |
| "loss": 1.3389, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9622680003092503e-05, | |
| "loss": 1.2762, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.961749119215287e-05, | |
| "loss": 1.2946, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.961226764234185e-05, | |
| "loss": 1.3227, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9607009372527172e-05, | |
| "loss": 1.3411, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9601716401701976e-05, | |
| "loss": 1.3325, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.959638874898475e-05, | |
| "loss": 1.4039, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9591026433619246e-05, | |
| "loss": 1.3013, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9585629474974413e-05, | |
| "loss": 1.3372, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9580197892544354e-05, | |
| "loss": 1.3755, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.957473170594822e-05, | |
| "loss": 1.374, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9569230934930157e-05, | |
| "loss": 1.329, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9563695599359233e-05, | |
| "loss": 1.3057, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9558125719229357e-05, | |
| "loss": 1.3024, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.955252131465923e-05, | |
| "loss": 1.2636, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9546882405892247e-05, | |
| "loss": 1.2914, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9541209013296435e-05, | |
| "loss": 1.3777, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9535501157364375e-05, | |
| "loss": 1.333, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.952975885871314e-05, | |
| "loss": 1.4056, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.952398213808421e-05, | |
| "loss": 1.3496, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.951817101634339e-05, | |
| "loss": 1.3138, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.951232551448075e-05, | |
| "loss": 1.2944, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9506445653610548e-05, | |
| "loss": 1.3374, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9500531454971136e-05, | |
| "loss": 1.3748, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9494582939924908e-05, | |
| "loss": 1.3424, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.94886001299582e-05, | |
| "loss": 1.313, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9482583046681238e-05, | |
| "loss": 1.3227, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9476531711828027e-05, | |
| "loss": 1.2002, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.94704461472563e-05, | |
| "loss": 1.3451, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9464326374947432e-05, | |
| "loss": 1.3672, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9458172417006347e-05, | |
| "loss": 1.3881, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9451984295661467e-05, | |
| "loss": 1.3065, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9445762033264595e-05, | |
| "loss": 1.36, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9439505652290867e-05, | |
| "loss": 1.3986, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.943321517533865e-05, | |
| "loss": 1.4003, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9426890625129466e-05, | |
| "loss": 1.2915, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.942053202450792e-05, | |
| "loss": 1.3677, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9414139396441602e-05, | |
| "loss": 1.3934, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9407712764021015e-05, | |
| "loss": 1.2412, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9401252150459478e-05, | |
| "loss": 1.2942, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9394757579093063e-05, | |
| "loss": 1.3166, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9388229073380495e-05, | |
| "loss": 1.313, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9381666656903068e-05, | |
| "loss": 1.2928, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9375070353364573e-05, | |
| "loss": 1.317, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9368440186591192e-05, | |
| "loss": 1.349, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9361776180531428e-05, | |
| "loss": 1.3727, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.935507835925601e-05, | |
| "loss": 1.3032, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.934834674695782e-05, | |
| "loss": 1.3728, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.934158136795178e-05, | |
| "loss": 1.2144, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9334782246674786e-05, | |
| "loss": 1.2674, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9327949407685617e-05, | |
| "loss": 1.3519, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9321082875664834e-05, | |
| "loss": 1.3787, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9314182675414704e-05, | |
| "loss": 1.3495, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9307248831859106e-05, | |
| "loss": 1.2048, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9300281370043433e-05, | |
| "loss": 1.3263, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9293280315134524e-05, | |
| "loss": 1.2988, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.928624569242054e-05, | |
| "loss": 1.2418, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.92791775273109e-05, | |
| "loss": 1.3163, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9272075845336177e-05, | |
| "loss": 1.2373, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9264940672148018e-05, | |
| "loss": 1.2492, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9257772033519032e-05, | |
| "loss": 1.2972, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9250569955342705e-05, | |
| "loss": 1.2825, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9243334463633315e-05, | |
| "loss": 1.3018, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.923606558452583e-05, | |
| "loss": 1.4347, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9228763344275813e-05, | |
| "loss": 1.301, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9221427769259333e-05, | |
| "loss": 1.2397, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.921405888597286e-05, | |
| "loss": 1.3608, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9206656721033178e-05, | |
| "loss": 1.359, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9199221301177287e-05, | |
| "loss": 1.323, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.919175265326231e-05, | |
| "loss": 1.3361, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9184250804265375e-05, | |
| "loss": 1.3126, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9176715781283556e-05, | |
| "loss": 1.2873, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9169147611533737e-05, | |
| "loss": 1.3034, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.916154632235254e-05, | |
| "loss": 1.3788, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.91539119411962e-05, | |
| "loss": 1.2423, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.914624449564051e-05, | |
| "loss": 1.3597, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.913854401338066e-05, | |
| "loss": 1.2726, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.91308105222312e-05, | |
| "loss": 1.3245, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9123044050125892e-05, | |
| "loss": 1.3731, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9115244625117627e-05, | |
| "loss": 1.2859, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9107412275378335e-05, | |
| "loss": 1.306, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.909954702919886e-05, | |
| "loss": 1.3568, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9091648914988873e-05, | |
| "loss": 1.2589, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9083717961276775e-05, | |
| "loss": 1.2778, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9075754196709574e-05, | |
| "loss": 1.2671, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9067757650052792e-05, | |
| "loss": 1.284, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.905972835019037e-05, | |
| "loss": 1.3116, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.905166632612455e-05, | |
| "loss": 1.3019, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9043571606975776e-05, | |
| "loss": 1.278, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.903544422198259e-05, | |
| "loss": 1.2977, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.902728420050152e-05, | |
| "loss": 1.2591, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9019091572006988e-05, | |
| "loss": 1.395, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9010866366091182e-05, | |
| "loss": 1.3454, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.900260861246397e-05, | |
| "loss": 1.3167, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.899431834095278e-05, | |
| "loss": 1.3569, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.89859955815025e-05, | |
| "loss": 1.3338, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8977640364175367e-05, | |
| "loss": 1.3153, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8969252719150847e-05, | |
| "loss": 1.3261, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.896083267672555e-05, | |
| "loss": 1.2942, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.89523802673131e-05, | |
| "loss": 1.2212, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.894389552144403e-05, | |
| "loss": 1.3297, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.893537846976568e-05, | |
| "loss": 1.2216, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.8926829143042075e-05, | |
| "loss": 1.3309, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8918247572153822e-05, | |
| "loss": 1.312, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8909633788098004e-05, | |
| "loss": 1.3446, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8900987821988038e-05, | |
| "loss": 1.3735, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.88923097050536e-05, | |
| "loss": 1.3151, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8883599468640498e-05, | |
| "loss": 1.3641, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.887485714421055e-05, | |
| "loss": 1.3251, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.886608276334148e-05, | |
| "loss": 1.3444, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8857276357726802e-05, | |
| "loss": 1.3449, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8848437959175703e-05, | |
| "loss": 1.3467, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8839567599612934e-05, | |
| "loss": 1.2495, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8830665311078694e-05, | |
| "loss": 1.3242, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8821731125728493e-05, | |
| "loss": 1.2344, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8812765075833076e-05, | |
| "loss": 1.2997, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8803767193778273e-05, | |
| "loss": 1.2249, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.879473751206489e-05, | |
| "loss": 1.2523, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.878567606330861e-05, | |
| "loss": 1.3366, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.8776582880239847e-05, | |
| "loss": 1.3649, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.876745799570364e-05, | |
| "loss": 1.3518, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8758301442659546e-05, | |
| "loss": 1.1984, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8749113254181498e-05, | |
| "loss": 1.397, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.873989346345771e-05, | |
| "loss": 1.2047, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8730642103790537e-05, | |
| "loss": 1.3858, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.872135920859636e-05, | |
| "loss": 1.2031, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.871204481140548e-05, | |
| "loss": 1.4161, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.870269894586197e-05, | |
| "loss": 1.3427, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8693321645723584e-05, | |
| "loss": 1.2259, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8683912944861606e-05, | |
| "loss": 1.3471, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.867447287726074e-05, | |
| "loss": 1.2364, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8665001477019006e-05, | |
| "loss": 1.289, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.865549877834758e-05, | |
| "loss": 1.2322, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.86459648155707e-05, | |
| "loss": 1.262, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8636399623125524e-05, | |
| "loss": 1.347, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8626803235562025e-05, | |
| "loss": 1.2308, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.8617175687542844e-05, | |
| "loss": 1.3258, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8607517013843172e-05, | |
| "loss": 1.1856, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.859782724935064e-05, | |
| "loss": 1.3437, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8588106429065175e-05, | |
| "loss": 1.2403, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8578354588098877e-05, | |
| "loss": 1.2893, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8568571761675893e-05, | |
| "loss": 1.2987, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8558757985132296e-05, | |
| "loss": 1.2255, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8548913293915954e-05, | |
| "loss": 1.3965, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8539037723586387e-05, | |
| "loss": 1.2476, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8529131309814667e-05, | |
| "loss": 1.2654, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.851919408838327e-05, | |
| "loss": 1.2373, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.850922609518595e-05, | |
| "loss": 1.2854, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8499227366227605e-05, | |
| "loss": 1.3365, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8489197937624162e-05, | |
| "loss": 1.2573, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8479137845602426e-05, | |
| "loss": 1.2512, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8469047126499967e-05, | |
| "loss": 1.3111, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8458925816764983e-05, | |
| "loss": 1.2592, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8448773952956164e-05, | |
| "loss": 1.3152, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8438591571742556e-05, | |
| "loss": 1.3005, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8428378709903447e-05, | |
| "loss": 1.1925, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8418135404328218e-05, | |
| "loss": 1.3367, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8407861692016215e-05, | |
| "loss": 1.1738, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.839755761007661e-05, | |
| "loss": 1.3063, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8387223195728276e-05, | |
| "loss": 1.279, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.837685848629965e-05, | |
| "loss": 1.2847, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.836646351922859e-05, | |
| "loss": 1.3267, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8356038332062258e-05, | |
| "loss": 1.3021, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8345582962456955e-05, | |
| "loss": 1.2832, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.833509744817802e-05, | |
| "loss": 1.2527, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8324581827099665e-05, | |
| "loss": 1.2557, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8314036137204853e-05, | |
| "loss": 1.3615, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.8303460416585155e-05, | |
| "loss": 1.3034, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.829285470344062e-05, | |
| "loss": 1.2433, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.828221903607962e-05, | |
| "loss": 1.3128, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.827155345291874e-05, | |
| "loss": 1.3725, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.826085799248261e-05, | |
| "loss": 1.2327, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.825013269340378e-05, | |
| "loss": 1.3472, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.823937759442258e-05, | |
| "loss": 1.3852, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8228592734386983e-05, | |
| "loss": 1.3841, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.821777815225245e-05, | |
| "loss": 1.3298, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8206933887081817e-05, | |
| "loss": 1.2682, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8196059978045117e-05, | |
| "loss": 1.256, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8185156464419467e-05, | |
| "loss": 1.2935, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.817422338558892e-05, | |
| "loss": 1.239, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8163260781044317e-05, | |
| "loss": 1.3453, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8152268690383146e-05, | |
| "loss": 1.288, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.814124715330941e-05, | |
| "loss": 1.2275, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8130196209633455e-05, | |
| "loss": 1.3109, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8119115899271863e-05, | |
| "loss": 1.4089, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.8108006262247282e-05, | |
| "loss": 1.3837, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.809686733868829e-05, | |
| "loss": 1.2909, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.808569916882925e-05, | |
| "loss": 1.2858, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8074501793010166e-05, | |
| "loss": 1.4052, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.806327525167653e-05, | |
| "loss": 1.2526, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.805201958537919e-05, | |
| "loss": 1.2896, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8040734834774186e-05, | |
| "loss": 1.3058, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8029421040622614e-05, | |
| "loss": 1.27, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8018078243790484e-05, | |
| "loss": 1.3524, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.8006706485248558e-05, | |
| "loss": 1.3095, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.799530580607221e-05, | |
| "loss": 1.2465, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7983876247441282e-05, | |
| "loss": 1.2959, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7972417850639916e-05, | |
| "loss": 1.3024, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.796093065705644e-05, | |
| "loss": 1.3681, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.794941470818318e-05, | |
| "loss": 1.4034, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7937870045616334e-05, | |
| "loss": 1.3187, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7926296711055818e-05, | |
| "loss": 1.2311, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.7914694746305107e-05, | |
| "loss": 1.3033, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.79030641932711e-05, | |
| "loss": 1.2861, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.789140509396394e-05, | |
| "loss": 1.335, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7879717490496896e-05, | |
| "loss": 1.2853, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7868001425086193e-05, | |
| "loss": 1.261, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.785625694005086e-05, | |
| "loss": 1.3158, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7844484077812575e-05, | |
| "loss": 1.2731, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7832682880895526e-05, | |
| "loss": 1.2974, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.782085339192624e-05, | |
| "loss": 1.2916, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7808995653633433e-05, | |
| "loss": 1.2562, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.779710970884787e-05, | |
| "loss": 1.2349, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.778519560050219e-05, | |
| "loss": 1.3608, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7773253371630762e-05, | |
| "loss": 1.3262, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7761283065369533e-05, | |
| "loss": 1.2279, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.774928472495586e-05, | |
| "loss": 1.3138, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7737258393728363e-05, | |
| "loss": 1.273, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7725204115126772e-05, | |
| "loss": 1.2071, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7713121932691755e-05, | |
| "loss": 1.2668, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.7701011890064778e-05, | |
| "loss": 1.1902, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.768887403098793e-05, | |
| "loss": 1.3153, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7676708399303784e-05, | |
| "loss": 1.2884, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.766451503895523e-05, | |
| "loss": 1.2719, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7652293993985304e-05, | |
| "loss": 1.2502, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.764004530853705e-05, | |
| "loss": 1.2796, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.762776902685335e-05, | |
| "loss": 1.2544, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.761546519327676e-05, | |
| "loss": 1.2752, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7603133852249357e-05, | |
| "loss": 1.2869, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7590775048312577e-05, | |
| "loss": 1.309, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.757838882610706e-05, | |
| "loss": 1.2327, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7565975230372473e-05, | |
| "loss": 1.2557, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7553534305947352e-05, | |
| "loss": 1.2468, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7541066097768965e-05, | |
| "loss": 1.2354, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7528570650873115e-05, | |
| "loss": 1.2976, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7516048010393993e-05, | |
| "loss": 1.304, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7503498221564026e-05, | |
| "loss": 1.2967, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.749092132971369e-05, | |
| "loss": 1.1766, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.7478317380271367e-05, | |
| "loss": 1.2731, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.746568641876316e-05, | |
| "loss": 1.2995, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7453028490812764e-05, | |
| "loss": 1.3078, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.744034364214125e-05, | |
| "loss": 1.2704, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.742763191856695e-05, | |
| "loss": 1.3429, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7414893366005263e-05, | |
| "loss": 1.2426, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7402128030468487e-05, | |
| "loss": 1.2139, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.738933595806567e-05, | |
| "loss": 1.2751, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.737651719500244e-05, | |
| "loss": 1.2601, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7363671787580818e-05, | |
| "loss": 1.2236, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.735079978219908e-05, | |
| "loss": 1.2135, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.733790122535157e-05, | |
| "loss": 1.2746, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7324976163628533e-05, | |
| "loss": 1.1513, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7312024643715958e-05, | |
| "loss": 1.2476, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7299046712395394e-05, | |
| "loss": 1.2501, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.72860424165438e-05, | |
| "loss": 1.2874, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.7273011803133356e-05, | |
| "loss": 1.2876, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.725995491923131e-05, | |
| "loss": 1.2822, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7246871811999793e-05, | |
| "loss": 1.2649, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.723376252869566e-05, | |
| "loss": 1.2315, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7220627116670314e-05, | |
| "loss": 1.2017, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7207465623369535e-05, | |
| "loss": 1.2219, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.719427809633332e-05, | |
| "loss": 1.319, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.718106458319568e-05, | |
| "loss": 1.3039, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7167825131684516e-05, | |
| "loss": 1.351, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7154559789621394e-05, | |
| "loss": 1.3211, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7141268604921414e-05, | |
| "loss": 1.2843, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.712795162559302e-05, | |
| "loss": 1.2754, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.711460889973782e-05, | |
| "loss": 1.2443, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7101240475550427e-05, | |
| "loss": 1.3022, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7087846401318273e-05, | |
| "loss": 1.2614, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7074426725421444e-05, | |
| "loss": 1.2902, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.706098149633249e-05, | |
| "loss": 1.2225, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7047510762616275e-05, | |
| "loss": 1.2784, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.7034014572929777e-05, | |
| "loss": 1.3031, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.702049297602193e-05, | |
| "loss": 1.2359, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.7006946020733426e-05, | |
| "loss": 1.261, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6993373755996572e-05, | |
| "loss": 1.3043, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6979776230835076e-05, | |
| "loss": 1.2774, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.69661534943639e-05, | |
| "loss": 1.3224, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.695250559578907e-05, | |
| "loss": 1.3853, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6938832584407494e-05, | |
| "loss": 1.3034, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6925134509606786e-05, | |
| "loss": 1.1885, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6911411420865097e-05, | |
| "loss": 1.309, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.689766336775093e-05, | |
| "loss": 1.3596, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6883890399922952e-05, | |
| "loss": 1.2906, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6870092567129837e-05, | |
| "loss": 1.2529, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6856269919210056e-05, | |
| "loss": 1.2574, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6842422506091727e-05, | |
| "loss": 1.2816, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6828550377792415e-05, | |
| "loss": 1.2797, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6814653584418966e-05, | |
| "loss": 1.2126, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.6800732176167296e-05, | |
| "loss": 1.1748, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6786786203322255e-05, | |
| "loss": 1.2303, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6772815716257414e-05, | |
| "loss": 1.3059, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.675882076543488e-05, | |
| "loss": 1.235, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6744801401405138e-05, | |
| "loss": 1.1984, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6730757674806858e-05, | |
| "loss": 1.1858, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6716689636366687e-05, | |
| "loss": 1.23, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6702597336899105e-05, | |
| "loss": 1.2098, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6688480827306224e-05, | |
| "loss": 1.2429, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.66743401585776e-05, | |
| "loss": 1.2814, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.666017538179005e-05, | |
| "loss": 1.1957, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.664598654810748e-05, | |
| "loss": 1.2829, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.663177370878068e-05, | |
| "loss": 1.301, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6617536915147156e-05, | |
| "loss": 1.2611, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6603276218630932e-05, | |
| "loss": 1.3957, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.6588991670742384e-05, | |
| "loss": 1.2579, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.657468332307803e-05, | |
| "loss": 1.2808, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.656035122732035e-05, | |
| "loss": 1.33, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.654599543523761e-05, | |
| "loss": 1.2839, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.653161599868367e-05, | |
| "loss": 1.2799, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6517212969597793e-05, | |
| "loss": 1.2972, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6502786400004454e-05, | |
| "loss": 1.3233, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6488336342013166e-05, | |
| "loss": 1.2354, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.647386284781828e-05, | |
| "loss": 1.2814, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6459365969698796e-05, | |
| "loss": 1.259, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.644484576001818e-05, | |
| "loss": 1.305, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6430302271224178e-05, | |
| "loss": 1.268, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.641573555584862e-05, | |
| "loss": 1.313, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6401145666507225e-05, | |
| "loss": 1.2821, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6386532655899418e-05, | |
| "loss": 1.3113, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6371896576808147e-05, | |
| "loss": 1.2607, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6357237482099682e-05, | |
| "loss": 1.2905, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.634255542472342e-05, | |
| "loss": 1.2955, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.632785045771171e-05, | |
| "loss": 1.2592, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.631312263417964e-05, | |
| "loss": 1.2158, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.6298372007324873e-05, | |
| "loss": 1.2978, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.628359863042742e-05, | |
| "loss": 1.3465, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6268802556849486e-05, | |
| "loss": 1.2226, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6253983840035243e-05, | |
| "loss": 1.1759, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.623914253351065e-05, | |
| "loss": 1.3524, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6224278690883277e-05, | |
| "loss": 1.2554, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.620939236584208e-05, | |
| "loss": 1.3068, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6194483612157232e-05, | |
| "loss": 1.1915, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.617955248367991e-05, | |
| "loss": 1.2537, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6164599034342122e-05, | |
| "loss": 1.3189, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6149623318156493e-05, | |
| "loss": 1.2137, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6134625389216076e-05, | |
| "loss": 1.2993, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6119605301694156e-05, | |
| "loss": 1.3054, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6104563109844063e-05, | |
| "loss": 1.3151, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.608949886799897e-05, | |
| "loss": 1.3125, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6074412630571685e-05, | |
| "loss": 1.2791, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.6059304452054475e-05, | |
| "loss": 1.2114, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.604417438701885e-05, | |
| "loss": 1.3071, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6029022490115383e-05, | |
| "loss": 1.2292, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.6013848816073507e-05, | |
| "loss": 1.2529, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5998653419701306e-05, | |
| "loss": 1.2233, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5983436355885333e-05, | |
| "loss": 1.2568, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5968197679590406e-05, | |
| "loss": 1.2929, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5952937445859396e-05, | |
| "loss": 1.2198, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.593765570981306e-05, | |
| "loss": 1.2318, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5922352526649803e-05, | |
| "loss": 1.1666, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.590702795164551e-05, | |
| "loss": 1.236, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.589168204015334e-05, | |
| "loss": 1.3213, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5876314847603506e-05, | |
| "loss": 1.2291, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5860926429503093e-05, | |
| "loss": 1.2891, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.584551684143586e-05, | |
| "loss": 1.1453, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5830086139062025e-05, | |
| "loss": 1.2638, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5814634378118085e-05, | |
| "loss": 1.2881, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.579916161441658e-05, | |
| "loss": 1.2684, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.5783667903845943e-05, | |
| "loss": 1.2407, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.576815330237024e-05, | |
| "loss": 1.2362, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5752617866029005e-05, | |
| "loss": 1.3086, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5737061650937033e-05, | |
| "loss": 1.3395, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5721484713284173e-05, | |
| "loss": 1.3664, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5705887109335124e-05, | |
| "loss": 1.2502, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.569026889542923e-05, | |
| "loss": 1.3197, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5674630127980278e-05, | |
| "loss": 1.2253, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5658970863476298e-05, | |
| "loss": 1.2318, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.564329115847936e-05, | |
| "loss": 1.2364, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.562759106962536e-05, | |
| "loss": 1.3453, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5611870653623826e-05, | |
| "loss": 1.1937, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5596129967257704e-05, | |
| "loss": 1.2392, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.558036906738316e-05, | |
| "loss": 1.1775, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5564588010929375e-05, | |
| "loss": 1.2464, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.554878685489834e-05, | |
| "loss": 1.1942, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.5532965656364627e-05, | |
| "loss": 1.2793, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.551712447247523e-05, | |
| "loss": 1.1996, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5501263360449303e-05, | |
| "loss": 1.2649, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.548538237757801e-05, | |
| "loss": 1.2232, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5469481581224274e-05, | |
| "loss": 1.238, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5453561028822573e-05, | |
| "loss": 1.2528, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5437620777878776e-05, | |
| "loss": 1.2723, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5421660885969875e-05, | |
| "loss": 1.0972, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5405681410743825e-05, | |
| "loss": 1.2458, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5389682409919308e-05, | |
| "loss": 1.2331, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5373663941285536e-05, | |
| "loss": 1.2441, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5357626062702035e-05, | |
| "loss": 1.3371, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5341568832098447e-05, | |
| "loss": 1.2398, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5325492307474314e-05, | |
| "loss": 1.2584, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.530939654689887e-05, | |
| "loss": 1.2923, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.529328160851082e-05, | |
| "loss": 1.2247, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5277147550518156e-05, | |
| "loss": 1.2918, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5260994431197916e-05, | |
| "loss": 1.289, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.5244822308896001e-05, | |
| "loss": 1.216, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.522863124202694e-05, | |
| "loss": 1.1868, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5212421289073703e-05, | |
| "loss": 1.2235, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5196192508587469e-05, | |
| "loss": 1.2868, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5179944959187423e-05, | |
| "loss": 1.2517, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5163678699560549e-05, | |
| "loss": 1.2021, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5147393788461411e-05, | |
| "loss": 1.1528, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5131090284711942e-05, | |
| "loss": 1.2786, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5114768247201235e-05, | |
| "loss": 1.2786, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5098427734885324e-05, | |
| "loss": 1.1867, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.508206880678698e-05, | |
| "loss": 1.2849, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.506569152199549e-05, | |
| "loss": 1.2415, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5049295939666445e-05, | |
| "loss": 1.2351, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.503288211902153e-05, | |
| "loss": 1.3409, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5016450119348309e-05, | |
| "loss": 1.2582, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.5000000000000002e-05, | |
| "loss": 1.2702, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.4983531820395287e-05, | |
| "loss": 1.2479, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.4967045640018074e-05, | |
| "loss": 1.1806, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4950541518417292e-05, | |
| "loss": 1.1697, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4934019515206675e-05, | |
| "loss": 1.298, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4917479690064545e-05, | |
| "loss": 1.3408, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4900922102733596e-05, | |
| "loss": 1.2368, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4884346813020685e-05, | |
| "loss": 1.2021, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4867753880796607e-05, | |
| "loss": 1.2987, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4851143365995886e-05, | |
| "loss": 1.2617, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4834515328616555e-05, | |
| "loss": 1.2645, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4817869828719933e-05, | |
| "loss": 1.2312, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4801206926430425e-05, | |
| "loss": 1.266, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4784526681935282e-05, | |
| "loss": 1.2838, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4767829155484407e-05, | |
| "loss": 1.3501, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4751114407390122e-05, | |
| "loss": 1.2235, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4734382498026946e-05, | |
| "loss": 1.1788, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4717633487831407e-05, | |
| "loss": 1.2717, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.470086743730178e-05, | |
| "loss": 1.2021, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.4684084406997903e-05, | |
| "loss": 1.3221, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.466728445754094e-05, | |
| "loss": 1.2328, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4650467649613168e-05, | |
| "loss": 1.2738, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4633634043957767e-05, | |
| "loss": 1.2546, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4616783701378573e-05, | |
| "loss": 1.226, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4599916682739898e-05, | |
| "loss": 1.2956, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4583033048966273e-05, | |
| "loss": 1.2365, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.456613286104225e-05, | |
| "loss": 1.2986, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4549216180012175e-05, | |
| "loss": 1.1944, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4532283066979967e-05, | |
| "loss": 1.3352, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4515333583108896e-05, | |
| "loss": 1.2509, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4498367789621369e-05, | |
| "loss": 1.2776, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4481385747798705e-05, | |
| "loss": 1.2736, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4464387518980905e-05, | |
| "loss": 1.2825, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.444737316456645e-05, | |
| "loss": 1.2081, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4430342746012049e-05, | |
| "loss": 1.2581, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.441329632483246e-05, | |
| "loss": 1.2339, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.4396233962600223e-05, | |
| "loss": 1.2929, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4379155720945464e-05, | |
| "loss": 1.3022, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4362061661555675e-05, | |
| "loss": 1.2035, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.434495184617547e-05, | |
| "loss": 1.1836, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4327826336606383e-05, | |
| "loss": 1.2184, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4310685194706626e-05, | |
| "loss": 1.2189, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.429352848239089e-05, | |
| "loss": 1.2939, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4276356261630096e-05, | |
| "loss": 1.3086, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4259168594451185e-05, | |
| "loss": 1.2081, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4241965542936894e-05, | |
| "loss": 1.1539, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4224747169225527e-05, | |
| "loss": 1.3016, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4207513535510736e-05, | |
| "loss": 1.2983, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4190264704041284e-05, | |
| "loss": 1.2656, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4173000737120839e-05, | |
| "loss": 1.2733, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.415572169710773e-05, | |
| "loss": 1.2133, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4138427646414746e-05, | |
| "loss": 1.3235, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4121118647508872e-05, | |
| "loss": 1.3324, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.4103794762911106e-05, | |
| "loss": 1.2721, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4086456055196208e-05, | |
| "loss": 1.2735, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.406910258699248e-05, | |
| "loss": 1.2134, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4051734420981537e-05, | |
| "loss": 1.2503, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4034351619898088e-05, | |
| "loss": 1.3149, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.4016954246529697e-05, | |
| "loss": 1.1784, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3999542363716571e-05, | |
| "loss": 1.2769, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3982116034351325e-05, | |
| "loss": 1.2285, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3964675321378756e-05, | |
| "loss": 1.266, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3947220287795611e-05, | |
| "loss": 1.2178, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3929750996650363e-05, | |
| "loss": 1.1795, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3912267511042994e-05, | |
| "loss": 1.272, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3894769894124748e-05, | |
| "loss": 1.2893, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3877258209097913e-05, | |
| "loss": 1.2912, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3859732519215596e-05, | |
| "loss": 1.2324, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3842192887781487e-05, | |
| "loss": 1.3233, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.3824639378149634e-05, | |
| "loss": 1.2629, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.380707205372421e-05, | |
| "loss": 1.2688, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3789490977959297e-05, | |
| "loss": 1.2448, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3771896214358642e-05, | |
| "loss": 1.203, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3754287826475428e-05, | |
| "loss": 1.2845, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3736665877912063e-05, | |
| "loss": 1.2954, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3719030432319925e-05, | |
| "loss": 1.1856, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3701381553399147e-05, | |
| "loss": 1.1456, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3683719304898382e-05, | |
| "loss": 1.2641, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3666043750614584e-05, | |
| "loss": 1.2101, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3648354954392758e-05, | |
| "loss": 1.2427, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.363065298012574e-05, | |
| "loss": 1.162, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3612937891753974e-05, | |
| "loss": 1.15, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3595209753265266e-05, | |
| "loss": 1.1676, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3577468628694566e-05, | |
| "loss": 1.2021, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3559714582123713e-05, | |
| "loss": 1.2588, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3541947677681246e-05, | |
| "loss": 1.2986, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3524167979542129e-05, | |
| "loss": 1.2277, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.3506375551927546e-05, | |
| "loss": 1.1577, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3488570459104656e-05, | |
| "loss": 1.3042, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3470752765386368e-05, | |
| "loss": 1.2314, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3452922535131108e-05, | |
| "loss": 1.2318, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3435079832742577e-05, | |
| "loss": 1.2922, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3417224722669534e-05, | |
| "loss": 1.2156, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.339935726940555e-05, | |
| "loss": 1.2158, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.338147753748879e-05, | |
| "loss": 1.1798, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3363585591501751e-05, | |
| "loss": 1.2187, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3345681496071068e-05, | |
| "loss": 1.2265, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3327765315867253e-05, | |
| "loss": 1.2686, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.330983711560446e-05, | |
| "loss": 1.2982, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3291896960040271e-05, | |
| "loss": 1.2733, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3273944913975452e-05, | |
| "loss": 1.2086, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.325598104225371e-05, | |
| "loss": 1.2497, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3238005409761468e-05, | |
| "loss": 1.3052, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3220018081427637e-05, | |
| "loss": 1.2156, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.3202019122223373e-05, | |
| "loss": 1.1905, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.318400859716183e-05, | |
| "loss": 1.2648, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3165986571297953e-05, | |
| "loss": 1.2634, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3147953109728224e-05, | |
| "loss": 1.2507, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3129908277590433e-05, | |
| "loss": 1.2471, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3111852140063435e-05, | |
| "loss": 1.2626, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3093784762366923e-05, | |
| "loss": 1.1367, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3075706209761201e-05, | |
| "loss": 1.1657, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.305761654754692e-05, | |
| "loss": 1.1931, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3039515841064867e-05, | |
| "loss": 1.3001, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3021404155695728e-05, | |
| "loss": 1.3272, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.3003281556859837e-05, | |
| "loss": 1.2923, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2985148110016947e-05, | |
| "loss": 1.2214, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2967003880666001e-05, | |
| "loss": 1.2456, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2948848934344885e-05, | |
| "loss": 1.1679, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2930683336630199e-05, | |
| "loss": 1.3211, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2912507153137009e-05, | |
| "loss": 1.1969, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.2894320449518624e-05, | |
| "loss": 1.3251, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2876123291466351e-05, | |
| "loss": 1.3732, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2857915744709255e-05, | |
| "loss": 1.1665, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.283969787501393e-05, | |
| "loss": 1.3549, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.282146974818426e-05, | |
| "loss": 1.2192, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.280323143006117e-05, | |
| "loss": 1.1423, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2784982986522399e-05, | |
| "loss": 1.2578, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2766724483482257e-05, | |
| "loss": 1.2487, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2748455986891406e-05, | |
| "loss": 1.2433, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2730177562736583e-05, | |
| "loss": 1.3278, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2711889277040392e-05, | |
| "loss": 1.2566, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2693591195861065e-05, | |
| "loss": 1.3571, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2675283385292212e-05, | |
| "loss": 1.2008, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2656965911462573e-05, | |
| "loss": 1.2369, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2638638840535808e-05, | |
| "loss": 1.1752, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2620302238710238e-05, | |
| "loss": 1.2576, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.260195617221861e-05, | |
| "loss": 1.2434, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.2583600707327854e-05, | |
| "loss": 1.258, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2565235910338852e-05, | |
| "loss": 1.2171, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2546861847586194e-05, | |
| "loss": 1.199, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2528478585437935e-05, | |
| "loss": 1.2697, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2510086190295358e-05, | |
| "loss": 1.2465, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2491684728592744e-05, | |
| "loss": 1.231, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2473274266797111e-05, | |
| "loss": 1.2877, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2454854871407993e-05, | |
| "loss": 1.2193, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2436426608957191e-05, | |
| "loss": 1.2549, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2417989546008533e-05, | |
| "loss": 1.2219, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.239954374915764e-05, | |
| "loss": 1.2328, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2381089285031666e-05, | |
| "loss": 1.2262, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2362626220289094e-05, | |
| "loss": 1.2424, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.234415462161945e-05, | |
| "loss": 1.2034, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2325674555743106e-05, | |
| "loss": 1.2005, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2307186089410999e-05, | |
| "loss": 1.2084, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2288689289404424e-05, | |
| "loss": 1.2174, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.2270184222534767e-05, | |
| "loss": 1.2185, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2251670955643277e-05, | |
| "loss": 1.1254, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2233149555600827e-05, | |
| "loss": 1.189, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2214620089307662e-05, | |
| "loss": 1.1525, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2196082623693169e-05, | |
| "loss": 1.2937, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2177537225715615e-05, | |
| "loss": 1.2662, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2158983962361938e-05, | |
| "loss": 1.2588, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2140422900647473e-05, | |
| "loss": 1.233, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2121854107615726e-05, | |
| "loss": 1.2085, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2103277650338128e-05, | |
| "loss": 1.1888, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.20846935959138e-05, | |
| "loss": 1.2574, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2066102011469295e-05, | |
| "loss": 1.2228, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2047502964158373e-05, | |
| "loss": 1.32, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2028896521161746e-05, | |
| "loss": 1.2702, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.2010282749686839e-05, | |
| "loss": 1.2662, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.199166171696755e-05, | |
| "loss": 1.1495, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.1973033490264e-05, | |
| "loss": 1.1424, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.1954398136862308e-05, | |
| "loss": 1.218, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1935755724074318e-05, | |
| "loss": 1.2007, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1917106319237386e-05, | |
| "loss": 1.338, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1898449989714116e-05, | |
| "loss": 1.2662, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1879786802892128e-05, | |
| "loss": 1.1688, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1861116826183807e-05, | |
| "loss": 1.2257, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.184244012702607e-05, | |
| "loss": 1.2185, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.182375677288011e-05, | |
| "loss": 1.2784, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1805066831231165e-05, | |
| "loss": 1.2824, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1786370369588257e-05, | |
| "loss": 1.1608, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1767667455483964e-05, | |
| "loss": 1.0812, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1748958156474176e-05, | |
| "loss": 1.1231, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1730242540137835e-05, | |
| "loss": 1.2081, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1711520674076711e-05, | |
| "loss": 1.2568, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.169279262591514e-05, | |
| "loss": 1.1896, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1674058463299798e-05, | |
| "loss": 1.2384, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1655318253899436e-05, | |
| "loss": 1.2358, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.1636572065404657e-05, | |
| "loss": 1.2163, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.161781996552765e-05, | |
| "loss": 1.1862, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1599062022001964e-05, | |
| "loss": 1.2108, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1580298302582249e-05, | |
| "loss": 1.1687, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1561528875044026e-05, | |
| "loss": 1.2398, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1542753807183429e-05, | |
| "loss": 1.2229, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1523973166816967e-05, | |
| "loss": 1.1675, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.150518702178127e-05, | |
| "loss": 1.1555, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1486395439932869e-05, | |
| "loss": 1.2433, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1467598489147911e-05, | |
| "loss": 1.2203, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1448796237321949e-05, | |
| "loss": 1.1818, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1429988752369681e-05, | |
| "loss": 1.2378, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1411176102224706e-05, | |
| "loss": 1.1441, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1392358354839286e-05, | |
| "loss": 1.2706, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1373535578184083e-05, | |
| "loss": 1.2184, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1354707840247936e-05, | |
| "loss": 1.1776, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.1335875209037598e-05, | |
| "loss": 1.169, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.13170377525775e-05, | |
| "loss": 1.1499, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1298195538909498e-05, | |
| "loss": 1.2536, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1279348636092634e-05, | |
| "loss": 1.1911, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1260497112202895e-05, | |
| "loss": 1.2602, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1241641035332944e-05, | |
| "loss": 1.2997, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1222780473591902e-05, | |
| "loss": 1.1709, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.120391549510508e-05, | |
| "loss": 1.2243, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1185046168013752e-05, | |
| "loss": 1.177, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1166172560474894e-05, | |
| "loss": 1.274, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1147294740660943e-05, | |
| "loss": 1.2195, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1128412776759556e-05, | |
| "loss": 1.2317, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.110952673697335e-05, | |
| "loss": 1.2189, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1090636689519673e-05, | |
| "loss": 1.3025, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1071742702630346e-05, | |
| "loss": 1.2653, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1052844844551416e-05, | |
| "loss": 1.0334, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.1033943183542914e-05, | |
| "loss": 1.2577, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.101503778787861e-05, | |
| "loss": 1.283, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.0996128725845764e-05, | |
| "loss": 1.2929, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.097721606574488e-05, | |
| "loss": 1.1483, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0958299875889451e-05, | |
| "loss": 1.202, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0939380224605727e-05, | |
| "loss": 1.1862, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.092045718023246e-05, | |
| "loss": 1.2189, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0901530811120655e-05, | |
| "loss": 1.2134, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0882601185633326e-05, | |
| "loss": 1.2334, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0863668372145255e-05, | |
| "loss": 1.3247, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0844732439042736e-05, | |
| "loss": 1.1771, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0825793454723325e-05, | |
| "loss": 1.1352, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.080685148759561e-05, | |
| "loss": 1.2442, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0787906606078942e-05, | |
| "loss": 1.1536, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.076895887860321e-05, | |
| "loss": 1.2705, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0750008373608571e-05, | |
| "loss": 1.2136, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.073105515954523e-05, | |
| "loss": 1.2627, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0712099304873161e-05, | |
| "loss": 1.3146, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0693140878061886e-05, | |
| "loss": 1.2138, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.0674179947590213e-05, | |
| "loss": 1.272, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0655216581945997e-05, | |
| "loss": 1.294, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0636250849625886e-05, | |
| "loss": 1.2523, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0617282819135077e-05, | |
| "loss": 1.2058, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0598312558987072e-05, | |
| "loss": 1.2699, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0579340137703422e-05, | |
| "loss": 1.2246, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0560365623813485e-05, | |
| "loss": 1.3123, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0541389085854177e-05, | |
| "loss": 1.2159, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0522410592369727e-05, | |
| "loss": 1.2068, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.050343021191143e-05, | |
| "loss": 1.2428, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.048444801303739e-05, | |
| "loss": 1.1799, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.046546406431228e-05, | |
| "loss": 1.172, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0446478434307103e-05, | |
| "loss": 1.2055, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0427491191598926e-05, | |
| "loss": 1.2507, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0408502404770641e-05, | |
| "loss": 1.2187, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0389512142410723e-05, | |
| "loss": 1.2031, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.037052047311297e-05, | |
| "loss": 1.1799, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.035152746547627e-05, | |
| "loss": 1.0826, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.0332533188104332e-05, | |
| "loss": 1.2806, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.031353770960547e-05, | |
| "loss": 1.248, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0294541098592323e-05, | |
| "loss": 1.2178, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0275543423681622e-05, | |
| "loss": 1.268, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0256544753493945e-05, | |
| "loss": 1.1488, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0237545156653461e-05, | |
| "loss": 1.2169, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0218544701787692e-05, | |
| "loss": 1.1846, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0199543457527247e-05, | |
| "loss": 1.2055, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0180541492505605e-05, | |
| "loss": 1.2382, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0161538875358833e-05, | |
| "loss": 1.2319, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.014253567472536e-05, | |
| "loss": 1.2129, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0123531959245721e-05, | |
| "loss": 1.2282, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0104527797562308e-05, | |
| "loss": 1.202, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0085523258319131e-05, | |
| "loss": 1.1793, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.006651841016156e-05, | |
| "loss": 1.0981, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0047513321736077e-05, | |
| "loss": 1.2196, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0028508061690044e-05, | |
| "loss": 1.2868, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.0009502698671427e-05, | |
| "loss": 1.2412, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.990497301328577e-06, | |
| "loss": 1.1552, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.971491938309963e-06, | |
| "loss": 1.2188, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.952486678263925e-06, | |
| "loss": 1.2852, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.933481589838442e-06, | |
| "loss": 1.2148, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.914476741680872e-06, | |
| "loss": 1.1657, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.895472202437695e-06, | |
| "loss": 1.2864, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.876468040754282e-06, | |
| "loss": 1.2121, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.857464325274643e-06, | |
| "loss": 1.2146, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.83846112464117e-06, | |
| "loss": 1.2107, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.819458507494395e-06, | |
| "loss": 1.2103, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.800456542472754e-06, | |
| "loss": 1.3071, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.781455298212313e-06, | |
| "loss": 1.0941, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.762454843346544e-06, | |
| "loss": 1.2122, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.743455246506059e-06, | |
| "loss": 1.2052, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.724456576318383e-06, | |
| "loss": 1.3103, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.70545890140768e-06, | |
| "loss": 1.0848, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 9.686462290394533e-06, | |
| "loss": 1.2722, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.667466811895668e-06, | |
| "loss": 1.2338, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.648472534523735e-06, | |
| "loss": 1.2152, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.629479526887032e-06, | |
| "loss": 1.1857, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.61048785758928e-06, | |
| "loss": 1.2306, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.59149759522936e-06, | |
| "loss": 1.1871, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.572508808401078e-06, | |
| "loss": 1.1799, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.553521565692902e-06, | |
| "loss": 1.202, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.534535935687723e-06, | |
| "loss": 1.2059, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.515551986962614e-06, | |
| "loss": 1.2164, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.496569788088575e-06, | |
| "loss": 1.2132, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.477589407630275e-06, | |
| "loss": 1.2383, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.458610914145826e-06, | |
| "loss": 1.212, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.439634376186519e-06, | |
| "loss": 1.3587, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.42065986229658e-06, | |
| "loss": 1.2435, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.401687441012928e-06, | |
| "loss": 1.2469, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.382717180864925e-06, | |
| "loss": 1.1605, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 9.363749150374115e-06, | |
| "loss": 1.1545, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.344783418054005e-06, | |
| "loss": 1.1999, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.32582005240979e-06, | |
| "loss": 1.2314, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.306859121938116e-06, | |
| "loss": 1.2782, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.287900695126844e-06, | |
| "loss": 1.1624, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.268944840454773e-06, | |
| "loss": 1.1685, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.249991626391429e-06, | |
| "loss": 1.2337, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.231041121396795e-06, | |
| "loss": 1.1369, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.212093393921061e-06, | |
| "loss": 1.1623, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.193148512404394e-06, | |
| "loss": 1.1005, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.174206545276678e-06, | |
| "loss": 1.1652, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.155267560957268e-06, | |
| "loss": 1.1361, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.136331627854745e-06, | |
| "loss": 1.2243, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.117398814366677e-06, | |
| "loss": 1.26, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.098469188879348e-06, | |
| "loss": 1.2799, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.079542819767544e-06, | |
| "loss": 1.1975, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.060619775394276e-06, | |
| "loss": 1.143, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 9.041700124110552e-06, | |
| "loss": 1.2919, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.022783934255126e-06, | |
| "loss": 1.1888, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.003871274154237e-06, | |
| "loss": 1.2185, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.984962212121391e-06, | |
| "loss": 1.2615, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.96605681645709e-06, | |
| "loss": 1.2452, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.947155155448588e-06, | |
| "loss": 1.2532, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.928257297369657e-06, | |
| "loss": 1.1422, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.909363310480329e-06, | |
| "loss": 1.2151, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.890473263026653e-06, | |
| "loss": 1.2735, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.87158722324045e-06, | |
| "loss": 1.1515, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.85270525933906e-06, | |
| "loss": 1.1761, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.833827439525109e-06, | |
| "loss": 1.1464, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.814953831986253e-06, | |
| "loss": 1.1732, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.796084504894923e-06, | |
| "loss": 1.1793, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.777219526408103e-06, | |
| "loss": 1.1665, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.758358964667061e-06, | |
| "loss": 1.2439, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.739502887797108e-06, | |
| "loss": 1.1755, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 8.720651363907364e-06, | |
| "loss": 1.3034, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.701804461090506e-06, | |
| "loss": 1.262, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.682962247422502e-06, | |
| "loss": 1.1896, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.664124790962407e-06, | |
| "loss": 1.2612, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.645292159752067e-06, | |
| "loss": 1.2589, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.626464421815919e-06, | |
| "loss": 1.2455, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.607641645160721e-06, | |
| "loss": 1.2321, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.588823897775296e-06, | |
| "loss": 1.1979, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.57001124763032e-06, | |
| "loss": 1.2727, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.551203762678056e-06, | |
| "loss": 1.1765, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.532401510852092e-06, | |
| "loss": 1.1837, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.513604560067135e-06, | |
| "loss": 1.2359, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.494812978218732e-06, | |
| "loss": 1.1295, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.476026833183036e-06, | |
| "loss": 1.2414, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.457246192816571e-06, | |
| "loss": 1.155, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.438471124955979e-06, | |
| "loss": 1.1973, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.419701697417753e-06, | |
| "loss": 1.2768, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 8.400937977998043e-06, | |
| "loss": 1.1379, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.382180034472353e-06, | |
| "loss": 1.2184, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.363427934595347e-06, | |
| "loss": 1.2427, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.344681746100569e-06, | |
| "loss": 1.2858, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.325941536700206e-06, | |
| "loss": 1.1733, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.30720737408486e-06, | |
| "loss": 1.2643, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.288479325923294e-06, | |
| "loss": 1.2195, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.269757459862169e-06, | |
| "loss": 1.1873, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.251041843525828e-06, | |
| "loss": 1.1861, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.232332544516038e-06, | |
| "loss": 1.2642, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.213629630411747e-06, | |
| "loss": 1.1918, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.194933168768842e-06, | |
| "loss": 1.1844, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.176243227119892e-06, | |
| "loss": 1.1905, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.157559872973932e-06, | |
| "loss": 1.1061, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.138883173816197e-06, | |
| "loss": 1.2447, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.120213197107877e-06, | |
| "loss": 1.1898, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.101550010285887e-06, | |
| "loss": 1.2071, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 8.082893680762619e-06, | |
| "loss": 1.2617, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.064244275925684e-06, | |
| "loss": 1.2059, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.045601863137694e-06, | |
| "loss": 1.3117, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.026966509736001e-06, | |
| "loss": 1.2692, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 8.008338283032452e-06, | |
| "loss": 1.2375, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.989717250313165e-06, | |
| "loss": 1.2351, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.971103478838256e-06, | |
| "loss": 1.2244, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.952497035841629e-06, | |
| "loss": 1.1868, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.93389798853071e-06, | |
| "loss": 1.3002, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.915306404086202e-06, | |
| "loss": 1.1335, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.896722349661872e-06, | |
| "loss": 1.2068, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.878145892384279e-06, | |
| "loss": 1.0868, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.859577099352528e-06, | |
| "loss": 1.1541, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.841016037638064e-06, | |
| "loss": 1.2877, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.822462774284389e-06, | |
| "loss": 1.1779, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.803917376306836e-06, | |
| "loss": 1.2062, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.785379910692343e-06, | |
| "loss": 1.2446, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 7.766850444399176e-06, | |
| "loss": 1.255, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.748329044356724e-06, | |
| "loss": 1.1769, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.72981577746524e-06, | |
| "loss": 1.2841, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.711310710595578e-06, | |
| "loss": 1.2994, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.692813910589001e-06, | |
| "loss": 1.1735, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.674325444256899e-06, | |
| "loss": 1.2689, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.655845378380551e-06, | |
| "loss": 1.2424, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.637373779710908e-06, | |
| "loss": 1.1596, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.618910714968336e-06, | |
| "loss": 1.1553, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.600456250842364e-06, | |
| "loss": 1.1199, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.582010453991468e-06, | |
| "loss": 1.2219, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.563573391042813e-06, | |
| "loss": 1.228, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.545145128592009e-06, | |
| "loss": 1.1553, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.526725733202894e-06, | |
| "loss": 1.1738, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.508315271407261e-06, | |
| "loss": 1.2284, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.489913809704643e-06, | |
| "loss": 1.215, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.471521414562071e-06, | |
| "loss": 1.1927, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 7.453138152413808e-06, | |
| "loss": 1.2281, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.434764089661151e-06, | |
| "loss": 1.2783, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.4163992926721505e-06, | |
| "loss": 1.1647, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.398043827781394e-06, | |
| "loss": 1.1822, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.379697761289763e-06, | |
| "loss": 1.2409, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.361361159464195e-06, | |
| "loss": 1.0965, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.34303408853743e-06, | |
| "loss": 1.1345, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.324716614707794e-06, | |
| "loss": 1.2526, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.306408804138937e-06, | |
| "loss": 1.2304, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.288110722959608e-06, | |
| "loss": 1.2212, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.269822437263423e-06, | |
| "loss": 1.1559, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.2515440131085975e-06, | |
| "loss": 1.2019, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.233275516517742e-06, | |
| "loss": 1.2414, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.215017013477607e-06, | |
| "loss": 1.1949, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.1967685699388345e-06, | |
| "loss": 1.2062, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.1785302518157415e-06, | |
| "loss": 1.1153, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.160302124986073e-06, | |
| "loss": 1.0821, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 7.1420842552907465e-06, | |
| "loss": 1.1893, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.123876708533655e-06, | |
| "loss": 1.1631, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.105679550481379e-06, | |
| "loss": 1.3004, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.087492846862992e-06, | |
| "loss": 1.1615, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.0693166633698055e-06, | |
| "loss": 1.1586, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.0511510656551175e-06, | |
| "loss": 1.2299, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.0329961193340014e-06, | |
| "loss": 1.2743, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 7.014851889983058e-06, | |
| "loss": 1.1953, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.9967184431401665e-06, | |
| "loss": 1.1268, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.978595844304272e-06, | |
| "loss": 1.2644, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.9604841589351346e-06, | |
| "loss": 1.1159, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.942383452453083e-06, | |
| "loss": 1.2025, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.924293790238803e-06, | |
| "loss": 1.2091, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.906215237633079e-06, | |
| "loss": 1.2076, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.888147859936569e-06, | |
| "loss": 1.1621, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.870091722409573e-06, | |
| "loss": 1.2223, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.852046890271778e-06, | |
| "loss": 1.2369, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 6.834013428702049e-06, | |
| "loss": 1.2712, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.815991402838176e-06, | |
| "loss": 1.1514, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.797980877776631e-06, | |
| "loss": 1.1875, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.779981918572364e-06, | |
| "loss": 1.1754, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.761994590238536e-06, | |
| "loss": 1.1287, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.744018957746295e-06, | |
| "loss": 1.1479, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.72605508602455e-06, | |
| "loss": 1.2319, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.7081030399597306e-06, | |
| "loss": 1.0982, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.690162884395541e-06, | |
| "loss": 1.1652, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.6722346841327515e-06, | |
| "loss": 1.1272, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.654318503928934e-06, | |
| "loss": 1.2393, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.636414408498249e-06, | |
| "loss": 1.2529, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.6185224625112155e-06, | |
| "loss": 1.1798, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.600642730594451e-06, | |
| "loss": 1.2632, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.582775277330467e-06, | |
| "loss": 1.1552, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.564920167257427e-06, | |
| "loss": 1.1897, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.5470774648688965e-06, | |
| "loss": 1.1429, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 6.529247234613634e-06, | |
| "loss": 1.2584, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.511429540895349e-06, | |
| "loss": 1.2007, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.4936244480724575e-06, | |
| "loss": 1.2732, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.475832020457875e-06, | |
| "loss": 1.2674, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.458052322318758e-06, | |
| "loss": 1.167, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.440285417876288e-06, | |
| "loss": 1.2319, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.42253137130544e-06, | |
| "loss": 1.2682, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.4047902467347355e-06, | |
| "loss": 1.2427, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.387062108246026e-06, | |
| "loss": 1.2025, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.369347019874264e-06, | |
| "loss": 1.1095, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.3516450456072465e-06, | |
| "loss": 1.1348, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.333956249385418e-06, | |
| "loss": 1.1869, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.316280695101621e-06, | |
| "loss": 1.0736, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.298618446600856e-06, | |
| "loss": 1.2294, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.28096956768008e-06, | |
| "loss": 1.1407, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.26333412208794e-06, | |
| "loss": 1.1574, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.245712173524572e-06, | |
| "loss": 1.2368, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 6.2281037856413625e-06, | |
| "loss": 1.232, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.210509022040705e-06, | |
| "loss": 1.1226, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.1929279462757905e-06, | |
| "loss": 1.1826, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.175360621850371e-06, | |
| "loss": 1.1734, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.157807112218517e-06, | |
| "loss": 1.1807, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.140267480784405e-06, | |
| "loss": 1.158, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.122741790902089e-06, | |
| "loss": 1.2017, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.105230105875254e-06, | |
| "loss": 1.164, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.08773248895701e-06, | |
| "loss": 1.1577, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.070249003349639e-06, | |
| "loss": 1.2334, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.052779712204391e-06, | |
| "loss": 1.2638, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.035324678621248e-06, | |
| "loss": 1.1633, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.017883965648676e-06, | |
| "loss": 1.2882, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 6.0004576362834295e-06, | |
| "loss": 1.1283, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.983045753470308e-06, | |
| "loss": 1.2374, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.965648380101916e-06, | |
| "loss": 1.1786, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.948265579018464e-06, | |
| "loss": 1.2134, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 5.930897413007523e-06, | |
| "loss": 1.1776, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.913543944803795e-06, | |
| "loss": 1.2078, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.896205237088894e-06, | |
| "loss": 1.1323, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.8788813524911324e-06, | |
| "loss": 1.1163, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.861572353585259e-06, | |
| "loss": 1.253, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.8442783028922725e-06, | |
| "loss": 1.2622, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.826999262879164e-06, | |
| "loss": 1.1634, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.809735295958718e-06, | |
| "loss": 1.1506, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.7924864644892675e-06, | |
| "loss": 1.1662, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.775252830774475e-06, | |
| "loss": 1.216, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.758034457063105e-06, | |
| "loss": 1.2356, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.740831405548819e-06, | |
| "loss": 1.2559, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.723643738369909e-06, | |
| "loss": 1.1466, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.706471517609111e-06, | |
| "loss": 1.1816, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.689314805293378e-06, | |
| "loss": 1.1695, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.672173663393618e-06, | |
| "loss": 1.2402, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.655048153824534e-06, | |
| "loss": 1.2497, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 5.637938338444325e-06, | |
| "loss": 1.1549, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.620844279054536e-06, | |
| "loss": 1.0983, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.60376603739978e-06, | |
| "loss": 1.1765, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.586703675167544e-06, | |
| "loss": 1.2167, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.569657253987952e-06, | |
| "loss": 1.1942, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.552626835433556e-06, | |
| "loss": 1.1483, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.535612481019098e-06, | |
| "loss": 1.2054, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.518614252201295e-06, | |
| "loss": 1.1733, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.5016322103786345e-06, | |
| "loss": 1.1337, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.484666416891109e-06, | |
| "loss": 1.2469, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.46771693302004e-06, | |
| "loss": 1.1667, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.45078381998783e-06, | |
| "loss": 1.1221, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.433867138957751e-06, | |
| "loss": 1.1597, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.416966951033731e-06, | |
| "loss": 1.252, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.400083317260103e-06, | |
| "loss": 1.2236, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.383216298621427e-06, | |
| "loss": 1.2796, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.366365956042236e-06, | |
| "loss": 1.1638, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 5.349532350386832e-06, | |
| "loss": 1.1897, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.3327155424590635e-06, | |
| "loss": 1.2024, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.3159155930021e-06, | |
| "loss": 1.2395, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.299132562698222e-06, | |
| "loss": 1.1615, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.282366512168595e-06, | |
| "loss": 1.1417, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.265617501973055e-06, | |
| "loss": 1.1685, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.248885592609885e-06, | |
| "loss": 1.1574, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.2321708445155985e-06, | |
| "loss": 1.1735, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.215473318064722e-06, | |
| "loss": 1.1609, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.198793073569577e-06, | |
| "loss": 1.2096, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.182130171280071e-06, | |
| "loss": 1.1439, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.165484671383445e-06, | |
| "loss": 1.2016, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.148856634004113e-06, | |
| "loss": 1.2194, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.1322461192033945e-06, | |
| "loss": 1.1392, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.115653186979318e-06, | |
| "loss": 1.2411, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.099077897266407e-06, | |
| "loss": 1.1637, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.082520309935459e-06, | |
| "loss": 1.1294, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 5.065980484793328e-06, | |
| "loss": 1.2772, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.04945848158271e-06, | |
| "loss": 1.1911, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.032954359981927e-06, | |
| "loss": 1.0489, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.016468179604712e-06, | |
| "loss": 1.2083, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 5.000000000000003e-06, | |
| "loss": 1.2155, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.9835498806516965e-06, | |
| "loss": 1.228, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.96711788097847e-06, | |
| "loss": 1.1028, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.95070406033356e-06, | |
| "loss": 1.1875, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.934308478004511e-06, | |
| "loss": 1.1926, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.91793119321302e-06, | |
| "loss": 1.2118, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.901572265114677e-06, | |
| "loss": 1.1967, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.885231752798767e-06, | |
| "loss": 1.2201, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.86890971528806e-06, | |
| "loss": 1.24, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.852606211538591e-06, | |
| "loss": 1.2671, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.836321300439454e-06, | |
| "loss": 1.219, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.82005504081258e-06, | |
| "loss": 1.2516, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.8038074914125345e-06, | |
| "loss": 1.1835, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.787578710926296e-06, | |
| "loss": 1.184, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.771368757973062e-06, | |
| "loss": 1.1205, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.755177691104004e-06, | |
| "loss": 1.1824, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.739005568802083e-06, | |
| "loss": 1.1991, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.722852449481848e-06, | |
| "loss": 1.2295, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.706718391489179e-06, | |
| "loss": 1.1957, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.690603453101134e-06, | |
| "loss": 1.2337, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.674507692525685e-06, | |
| "loss": 1.1564, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.658431167901553e-06, | |
| "loss": 1.1294, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.642373937297968e-06, | |
| "loss": 1.2489, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.626336058714468e-06, | |
| "loss": 1.2655, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.610317590080694e-06, | |
| "loss": 1.2224, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.594318589256177e-06, | |
| "loss": 1.1202, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.578339114030127e-06, | |
| "loss": 1.1434, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.562379222121224e-06, | |
| "loss": 1.1676, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.546438971177429e-06, | |
| "loss": 1.1515, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.530518418775734e-06, | |
| "loss": 1.1223, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.514617622421993e-06, | |
| "loss": 1.2414, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.4987366395506996e-06, | |
| "loss": 1.1827, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.482875527524774e-06, | |
| "loss": 1.1472, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.467034343635377e-06, | |
| "loss": 1.2123, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.451213145101664e-06, | |
| "loss": 1.1437, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.435411989070624e-06, | |
| "loss": 1.0953, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.419630932616841e-06, | |
| "loss": 1.2336, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.403870032742299e-06, | |
| "loss": 1.2267, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.388129346376177e-06, | |
| "loss": 1.2801, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.372408930374642e-06, | |
| "loss": 1.1538, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.356708841520643e-06, | |
| "loss": 1.2236, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.341029136523701e-06, | |
| "loss": 1.2342, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.325369872019727e-06, | |
| "loss": 1.1724, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.309731104570775e-06, | |
| "loss": 1.2231, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.2941128906648786e-06, | |
| "loss": 1.0791, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.27851528671583e-06, | |
| "loss": 1.2699, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.262938349062967e-06, | |
| "loss": 1.2006, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.247382133971002e-06, | |
| "loss": 1.0275, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.231846697629763e-06, | |
| "loss": 1.2236, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2163320961540585e-06, | |
| "loss": 1.2633, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.200838385583417e-06, | |
| "loss": 1.2254, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1853656218819175e-06, | |
| "loss": 1.1661, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.169913860937975e-06, | |
| "loss": 1.1264, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.154483158564143e-06, | |
| "loss": 1.0827, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.13907357049691e-06, | |
| "loss": 1.2385, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.123685152396498e-06, | |
| "loss": 1.2062, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.1083179598466625e-06, | |
| "loss": 1.1504, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.092972048354491e-06, | |
| "loss": 1.1717, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.077647473350201e-06, | |
| "loss": 1.0936, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.062344290186947e-06, | |
| "loss": 1.0833, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.047062554140605e-06, | |
| "loss": 1.1854, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.0318023204096015e-06, | |
| "loss": 1.1432, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.0165636441146676e-06, | |
| "loss": 1.2474, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.001346580298694e-06, | |
| "loss": 1.0607, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.986151183926494e-06, | |
| "loss": 1.2328, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 3.970977509884617e-06, | |
| "loss": 1.1857, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.955825612981156e-06, | |
| "loss": 1.1569, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.940695547945528e-06, | |
| "loss": 1.2401, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.925587369428316e-06, | |
| "loss": 1.137, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.910501132001031e-06, | |
| "loss": 1.2371, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.895436890155936e-06, | |
| "loss": 1.2359, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.880394698305846e-06, | |
| "loss": 1.1864, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8653746107839285e-06, | |
| "loss": 1.0934, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.850376681843511e-06, | |
| "loss": 1.1536, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8354009656578775e-06, | |
| "loss": 1.1341, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.820447516320093e-06, | |
| "loss": 1.2344, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.8055163878427703e-06, | |
| "loss": 1.1653, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7906076341579235e-06, | |
| "loss": 1.1705, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7757213091167243e-06, | |
| "loss": 1.2219, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7608574664893504e-06, | |
| "loss": 1.1946, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.746016159964764e-06, | |
| "loss": 1.1176, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.731197443150515e-06, | |
| "loss": 1.2197, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 3.7164013695725798e-06, | |
| "loss": 1.2031, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.70162799267513e-06, | |
| "loss": 1.2194, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6868773658203617e-06, | |
| "loss": 1.1545, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6721495422882948e-06, | |
| "loss": 1.1085, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.657444575276583e-06, | |
| "loss": 1.2488, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6427625179003223e-06, | |
| "loss": 1.3018, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6281034231918554e-06, | |
| "loss": 1.1583, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.6134673441005864e-06, | |
| "loss": 1.236, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5988543334927773e-06, | |
| "loss": 1.1291, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.584264444151384e-06, | |
| "loss": 1.2305, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5696977287758205e-06, | |
| "loss": 1.1356, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5551542399818204e-06, | |
| "loss": 1.2003, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5406340303012075e-06, | |
| "loss": 1.135, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.5261371521817247e-06, | |
| "loss": 1.2017, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.511663657986837e-06, | |
| "loss": 1.1771, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.497213599995549e-06, | |
| "loss": 1.2021, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.4827870304022116e-06, | |
| "loss": 1.158, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 3.468384001316333e-06, | |
| "loss": 1.0989, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.454004564762393e-06, | |
| "loss": 1.1485, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4396487726796545e-06, | |
| "loss": 1.1208, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.4253166769219746e-06, | |
| "loss": 1.1925, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.411008329257618e-06, | |
| "loss": 1.2145, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.396723781369067e-06, | |
| "loss": 1.1237, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3824630848528505e-06, | |
| "loss": 1.1729, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.368226291219323e-06, | |
| "loss": 1.0858, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3540134518925226e-06, | |
| "loss": 1.1509, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.33982461820995e-06, | |
| "loss": 1.1558, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.3256598414224027e-06, | |
| "loss": 1.1901, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.311519172693778e-06, | |
| "loss": 1.2092, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.297402663100897e-06, | |
| "loss": 1.2764, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.2833103636333176e-06, | |
| "loss": 1.1672, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.269242325193147e-06, | |
| "loss": 1.1626, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.255198598594862e-06, | |
| "loss": 1.2014, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.241179234565123e-06, | |
| "loss": 1.1559, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 3.2271842837425917e-06, | |
| "loss": 1.1703, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.2132137966777476e-06, | |
| "loss": 1.1252, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.199267823832705e-06, | |
| "loss": 1.2093, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.185346415581041e-06, | |
| "loss": 1.2315, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.171449622207584e-06, | |
| "loss": 1.1437, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.157577493908274e-06, | |
| "loss": 1.1403, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1437300807899462e-06, | |
| "loss": 1.1996, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.129907432870166e-06, | |
| "loss": 1.1522, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.116109600077052e-06, | |
| "loss": 1.2124, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.1023366322490733e-06, | |
| "loss": 1.1728, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0885885791349045e-06, | |
| "loss": 1.1787, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0748654903932164e-06, | |
| "loss": 1.153, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0611674155925085e-06, | |
| "loss": 1.2085, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.047494404210931e-06, | |
| "loss": 1.1711, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0338465056361e-06, | |
| "loss": 1.2137, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.0202237691649263e-06, | |
| "loss": 1.0411, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.00662624400343e-06, | |
| "loss": 1.1736, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 2.9930539792665767e-06, | |
| "loss": 1.1268, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9795070239780734e-06, | |
| "loss": 1.1023, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.965985427070226e-06, | |
| "loss": 1.2637, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9524892373837253e-06, | |
| "loss": 1.2217, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.939018503667511e-06, | |
| "loss": 1.1542, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9255732745785627e-06, | |
| "loss": 1.1, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.9121535986817284e-06, | |
| "loss": 1.1416, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8987595244495736e-06, | |
| "loss": 1.1453, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8853911002621803e-06, | |
| "loss": 1.182, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8720483744069817e-06, | |
| "loss": 1.2427, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8587313950785876e-06, | |
| "loss": 1.1616, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8454402103786096e-06, | |
| "loss": 1.1386, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8321748683154893e-06, | |
| "loss": 1.1982, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.8189354168043183e-06, | |
| "loss": 1.1384, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.805721903666685e-06, | |
| "loss": 1.2461, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7925343766304637e-06, | |
| "loss": 1.117, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7793728833296906e-06, | |
| "loss": 1.1167, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 2.7662374713043414e-06, | |
| "loss": 1.1186, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.753128188000208e-06, | |
| "loss": 1.1347, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.740045080768694e-06, | |
| "loss": 1.1603, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7269881968666446e-06, | |
| "loss": 1.1694, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.713957583456203e-06, | |
| "loss": 1.1141, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.7009532876046084e-06, | |
| "loss": 1.1761, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6879753562840473e-06, | |
| "loss": 1.148, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.675023836371471e-06, | |
| "loss": 1.194, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.662098774648434e-06, | |
| "loss": 1.1904, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6492002178009224e-06, | |
| "loss": 1.1511, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6363282124191847e-06, | |
| "loss": 1.1651, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.6234828049975645e-06, | |
| "loss": 1.1942, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.61066404193433e-06, | |
| "loss": 1.1863, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.597871969531518e-06, | |
| "loss": 1.1301, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.5851066339947397e-06, | |
| "loss": 1.1927, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.572368081433049e-06, | |
| "loss": 1.2208, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.559656357858753e-06, | |
| "loss": 1.2806, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.546971509187238e-06, | |
| "loss": 1.0795, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.534313581236839e-06, | |
| "loss": 1.2085, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.521682619728637e-06, | |
| "loss": 1.1635, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.5090786702863126e-06, | |
| "loss": 1.1319, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.496501778435977e-06, | |
| "loss": 1.1842, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4839519896060094e-06, | |
| "loss": 1.042, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4714293491268904e-06, | |
| "loss": 1.1816, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4589339022310386e-06, | |
| "loss": 1.1486, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4464656940526508e-06, | |
| "loss": 1.1069, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4340247696275297e-06, | |
| "loss": 1.1562, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.4216111738929426e-06, | |
| "loss": 1.1931, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.409224951687421e-06, | |
| "loss": 1.0961, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.396866147750644e-06, | |
| "loss": 1.1836, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3845348067232455e-06, | |
| "loss": 1.1885, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.372230973146652e-06, | |
| "loss": 1.0612, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3599546914629534e-06, | |
| "loss": 1.1892, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3477060060146973e-06, | |
| "loss": 1.2406, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.3354849610447728e-06, | |
| "loss": 1.204, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.323291600696217e-06, | |
| "loss": 1.1617, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.3111259690120726e-06, | |
| "loss": 1.0931, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.298988109935227e-06, | |
| "loss": 1.1107, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.286878067308248e-06, | |
| "loss": 1.2222, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.274795884873231e-06, | |
| "loss": 1.2072, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2627416062716366e-06, | |
| "loss": 1.0869, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.250715275044144e-06, | |
| "loss": 1.1507, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2387169346304683e-06, | |
| "loss": 1.2378, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2267466283692386e-06, | |
| "loss": 1.1628, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.214804399497815e-06, | |
| "loss": 1.1527, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.2028902911521324e-06, | |
| "loss": 1.1627, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.191004346366571e-06, | |
| "loss": 1.1561, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.179146608073763e-06, | |
| "loss": 1.2045, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1673171191044764e-06, | |
| "loss": 1.1289, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1555159221874267e-06, | |
| "loss": 1.1844, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.143743059949144e-06, | |
| "loss": 1.1313, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.1319985749138107e-06, | |
| "loss": 1.1146, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1202825095031077e-06, | |
| "loss": 1.2057, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.1085949060360654e-06, | |
| "loss": 1.0959, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0969358067289035e-06, | |
| "loss": 1.1209, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.085305253694895e-06, | |
| "loss": 1.161, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.073703288944183e-06, | |
| "loss": 1.2207, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0621299543836692e-06, | |
| "loss": 1.1743, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0505852918168244e-06, | |
| "loss": 1.2421, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.0390693429435626e-06, | |
| "loss": 1.1522, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.027582149360088e-06, | |
| "loss": 1.084, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.016123752558722e-06, | |
| "loss": 1.177, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.004694193927791e-06, | |
| "loss": 1.1559, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9932935147514445e-06, | |
| "loss": 1.195, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9819217562095184e-06, | |
| "loss": 1.113, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.970578959377388e-06, | |
| "loss": 1.0715, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.95926516522582e-06, | |
| "loss": 1.1071, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9479804146208147e-06, | |
| "loss": 1.1358, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.9367247483234743e-06, | |
| "loss": 1.155, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.9254982069898386e-06, | |
| "loss": 1.112, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.914300831170751e-06, | |
| "loss": 1.1614, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.903132661311714e-06, | |
| "loss": 1.2276, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.891993737752722e-06, | |
| "loss": 1.1827, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8808841007281375e-06, | |
| "loss": 1.1267, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8698037903665479e-06, | |
| "loss": 1.1975, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8587528466905925e-06, | |
| "loss": 1.2221, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8477313096168525e-06, | |
| "loss": 1.0311, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8367392189556843e-06, | |
| "loss": 1.1872, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8257766144110823e-06, | |
| "loss": 1.2141, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.814843535580536e-06, | |
| "loss": 1.1558, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8039400219548876e-06, | |
| "loss": 1.18, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.793066112918186e-06, | |
| "loss": 1.1717, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7822218477475496e-06, | |
| "loss": 1.1987, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7714072656130199e-06, | |
| "loss": 1.1674, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7606224055774202e-06, | |
| "loss": 1.1691, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.7498673065962235e-06, | |
| "loss": 1.0886, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7391420075173915e-06, | |
| "loss": 0.9991, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7284465470812605e-06, | |
| "loss": 1.209, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7177809639203834e-06, | |
| "loss": 1.1752, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.7071452965593839e-06, | |
| "loss": 1.2009, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6965395834148468e-06, | |
| "loss": 1.2222, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.685963862795149e-06, | |
| "loss": 1.1312, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.675418172900336e-06, | |
| "loss": 1.0991, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6649025518219809e-06, | |
| "loss": 1.1625, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6544170375430458e-06, | |
| "loss": 1.1371, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6439616679377445e-06, | |
| "loss": 1.2432, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6335364807714104e-06, | |
| "loss": 1.1782, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6231415137003536e-06, | |
| "loss": 1.1645, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6127768042717252e-06, | |
| "loss": 1.1696, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.6024423899233953e-06, | |
| "loss": 1.0632, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5921383079837882e-06, | |
| "loss": 1.0695, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5818645956717827e-06, | |
| "loss": 1.152, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.5716212900965565e-06, | |
| "loss": 1.1759, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.561408428257446e-06, | |
| "loss": 1.1607, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5512260470438422e-06, | |
| "loss": 1.1307, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5410741832350173e-06, | |
| "loss": 1.1464, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5309528735000335e-06, | |
| "loss": 1.2201, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5208621543975766e-06, | |
| "loss": 1.1855, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.5108020623758413e-06, | |
| "loss": 1.2273, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.500772633772396e-06, | |
| "loss": 1.241, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4907739048140524e-06, | |
| "loss": 1.1832, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4808059116167306e-06, | |
| "loss": 1.1978, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.470868690185332e-06, | |
| "loss": 1.1018, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4609622764136155e-06, | |
| "loss": 1.1125, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4510867060840483e-06, | |
| "loss": 1.1852, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4412420148677052e-06, | |
| "loss": 1.1264, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4314282383241097e-06, | |
| "loss": 1.169, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.4216454119011236e-06, | |
| "loss": 1.1919, | |
| "step": 1421 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.411893570934827e-06, | |
| "loss": 1.1465, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.40217275064936e-06, | |
| "loss": 1.1088, | |
| "step": 1423 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.392482986156829e-06, | |
| "loss": 1.1623, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3828243124571594e-06, | |
| "loss": 1.1215, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3731967644379761e-06, | |
| "loss": 1.1838, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3636003768744754e-06, | |
| "loss": 1.1409, | |
| "step": 1427 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.354035184429302e-06, | |
| "loss": 1.1206, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3445012216524223e-06, | |
| "loss": 1.1297, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.334998522980997e-06, | |
| "loss": 1.2138, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.325527122739262e-06, | |
| "loss": 1.1751, | |
| "step": 1431 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3160870551383975e-06, | |
| "loss": 1.149, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.3066783542764194e-06, | |
| "loss": 1.1542, | |
| "step": 1433 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2973010541380315e-06, | |
| "loss": 1.1354, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.287955188594522e-06, | |
| "loss": 1.0665, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2786407914036425e-06, | |
| "loss": 1.1461, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2693578962094666e-06, | |
| "loss": 1.2177, | |
| "step": 1437 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2601065365422915e-06, | |
| "loss": 1.1602, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2508867458185037e-06, | |
| "loss": 1.092, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.2416985573404571e-06, | |
| "loss": 1.1138, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.232542004296362e-06, | |
| "loss": 1.1082, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2234171197601563e-06, | |
| "loss": 1.1529, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2143239366913906e-06, | |
| "loss": 1.2283, | |
| "step": 1443 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.2052624879351105e-06, | |
| "loss": 1.1514, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.196232806221731e-06, | |
| "loss": 1.1593, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1872349241669245e-06, | |
| "loss": 1.0892, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1782688742715098e-06, | |
| "loss": 1.1828, | |
| "step": 1447 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1693346889213108e-06, | |
| "loss": 1.2005, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1604324003870637e-06, | |
| "loss": 1.1543, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1515620408242989e-06, | |
| "loss": 1.1349, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1427236422731991e-06, | |
| "loss": 1.1508, | |
| "step": 1451 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1339172366585216e-06, | |
| "loss": 1.1577, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1251428557894516e-06, | |
| "loss": 1.2336, | |
| "step": 1453 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.1164005313595038e-06, | |
| "loss": 1.1519, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.107690294946402e-06, | |
| "loss": 1.1119, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.0990121780119668e-06, | |
| "loss": 1.1607, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.0903662119020008e-06, | |
| "loss": 1.0861, | |
| "step": 1457 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0817524278461777e-06, | |
| "loss": 1.2109, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.073170856957928e-06, | |
| "loss": 1.1334, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.064621530234322e-06, | |
| "loss": 1.1235, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0561044785559737e-06, | |
| "loss": 1.1321, | |
| "step": 1461 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0476197326869043e-06, | |
| "loss": 1.1748, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0391673232744504e-06, | |
| "loss": 1.2237, | |
| "step": 1463 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0307472808491547e-06, | |
| "loss": 1.1894, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0223596358246346e-06, | |
| "loss": 1.2481, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0140044184975017e-06, | |
| "loss": 1.188, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.0056816590472206e-06, | |
| "loss": 1.2523, | |
| "step": 1467 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.973913875360331e-07, | |
| "loss": 1.1979, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.89133633908821e-07, | |
| "loss": 1.1918, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.809084279930159e-07, | |
| "loss": 1.159, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.727157994984815e-07, | |
| "loss": 1.1036, | |
| "step": 1471 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.64555778017413e-07, | |
| "loss": 1.1067, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.564283930242258e-07, | |
| "loss": 1.2142, | |
| "step": 1473 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 9.4833367387545e-07, | |
| "loss": 1.2832, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.402716498096331e-07, | |
| "loss": 1.1395, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.32242349947211e-07, | |
| "loss": 1.1718, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.242458032904311e-07, | |
| "loss": 1.1315, | |
| "step": 1477 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.162820387232285e-07, | |
| "loss": 1.1, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.083510850111277e-07, | |
| "loss": 1.2252, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 9.004529708011455e-07, | |
| "loss": 1.2342, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.925877246216686e-07, | |
| "loss": 1.1885, | |
| "step": 1481 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.847553748823745e-07, | |
| "loss": 1.2166, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.769559498741109e-07, | |
| "loss": 1.131, | |
| "step": 1483 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.691894777688004e-07, | |
| "loss": 1.1538, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.614559866193395e-07, | |
| "loss": 1.1186, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.537555043594936e-07, | |
| "loss": 1.0496, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.460880588038001e-07, | |
| "loss": 1.1004, | |
| "step": 1487 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.384536776474639e-07, | |
| "loss": 1.1879, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.308523884662656e-07, | |
| "loss": 1.2293, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.232842187164469e-07, | |
| "loss": 1.1442, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 8.157491957346275e-07, | |
| "loss": 1.1702, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.08247346737695e-07, | |
| "loss": 1.2673, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.007786988227139e-07, | |
| "loss": 1.1755, | |
| "step": 1493 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.93343278966825e-07, | |
| "loss": 1.0932, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.859411140271422e-07, | |
| "loss": 1.2368, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.785722307406685e-07, | |
| "loss": 1.2311, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.71236655724188e-07, | |
| "loss": 1.1807, | |
| "step": 1497 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.639344154741713e-07, | |
| "loss": 1.2103, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.566655363666864e-07, | |
| "loss": 1.1467, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.494300446572978e-07, | |
| "loss": 1.1643, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.422279664809706e-07, | |
| "loss": 1.2261, | |
| "step": 1501 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.350593278519824e-07, | |
| "loss": 1.0772, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.279241546638239e-07, | |
| "loss": 1.218, | |
| "step": 1503 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.208224726891044e-07, | |
| "loss": 1.1608, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.137543075794651e-07, | |
| "loss": 1.1196, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 7.0671968486548e-07, | |
| "loss": 1.2686, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 6.997186299565661e-07, | |
| "loss": 1.1473, | |
| "step": 1507 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 6.927511681408971e-07, | |
| "loss": 1.1463, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.858173245852962e-07, | |
| "loss": 1.1507, | |
| "step": 1509 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.789171243351678e-07, | |
| "loss": 1.1216, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.72050592314385e-07, | |
| "loss": 1.1535, | |
| "step": 1511 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.652177533252146e-07, | |
| "loss": 1.1821, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.584186320482222e-07, | |
| "loss": 1.0726, | |
| "step": 1513 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.516532530421826e-07, | |
| "loss": 1.1247, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.449216407439906e-07, | |
| "loss": 1.1563, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.382238194685752e-07, | |
| "loss": 1.129, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.315598134088119e-07, | |
| "loss": 1.138, | |
| "step": 1517 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.249296466354305e-07, | |
| "loss": 1.1219, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.18333343096933e-07, | |
| "loss": 1.1842, | |
| "step": 1519 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.117709266195083e-07, | |
| "loss": 1.0654, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 6.052424209069385e-07, | |
| "loss": 1.1573, | |
| "step": 1521 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.987478495405252e-07, | |
| "loss": 1.1874, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.922872359789878e-07, | |
| "loss": 1.1057, | |
| "step": 1523 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.858606035583992e-07, | |
| "loss": 1.2166, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 5.794679754920796e-07, | |
| "loss": 1.1399, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.731093748705341e-07, | |
| "loss": 1.1572, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.667848246613528e-07, | |
| "loss": 1.218, | |
| "step": 1527 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.604943477091351e-07, | |
| "loss": 1.2097, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.54237966735407e-07, | |
| "loss": 1.2243, | |
| "step": 1529 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.480157043385359e-07, | |
| "loss": 1.1342, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.418275829936537e-07, | |
| "loss": 1.1797, | |
| "step": 1531 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.356736250525718e-07, | |
| "loss": 1.2628, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.295538527437027e-07, | |
| "loss": 1.184, | |
| "step": 1533 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.234682881719766e-07, | |
| "loss": 1.1458, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.174169533187634e-07, | |
| "loss": 1.2195, | |
| "step": 1535 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.113998700418e-07, | |
| "loss": 1.1694, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 5.05417060075094e-07, | |
| "loss": 1.1742, | |
| "step": 1537 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.994685450288672e-07, | |
| "loss": 1.2063, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.935543463894554e-07, | |
| "loss": 1.2757, | |
| "step": 1539 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.876744855192516e-07, | |
| "loss": 1.1961, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.818289836566126e-07, | |
| "loss": 1.1616, | |
| "step": 1541 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.7601786191579293e-07, | |
| "loss": 1.2033, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 4.7024114128686017e-07, | |
| "loss": 1.1063, | |
| "step": 1543 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.6449884263562564e-07, | |
| "loss": 1.1181, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.587909867035678e-07, | |
| "loss": 1.1713, | |
| "step": 1545 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.531175941077537e-07, | |
| "loss": 1.0918, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.4747868534077197e-07, | |
| "loss": 1.1389, | |
| "step": 1547 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.4187428077064443e-07, | |
| "loss": 1.1615, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.363044006407724e-07, | |
| "loss": 1.0964, | |
| "step": 1549 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.3076906506984595e-07, | |
| "loss": 1.1582, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.252682940517805e-07, | |
| "loss": 1.1466, | |
| "step": 1551 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.198021074556469e-07, | |
| "loss": 1.19, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.1437052502558693e-07, | |
| "loss": 1.1323, | |
| "step": 1553 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.0897356638075815e-07, | |
| "loss": 1.2179, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.0361125101525124e-07, | |
| "loss": 1.093, | |
| "step": 1555 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.9828359829802263e-07, | |
| "loss": 1.1905, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.92990627472829e-07, | |
| "loss": 1.1468, | |
| "step": 1557 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.8773235765815155e-07, | |
| "loss": 1.2172, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.8250880784713064e-07, | |
| "loss": 1.1851, | |
| "step": 1559 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.773199969074959e-07, | |
| "loss": 1.1777, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.721659435814995e-07, | |
| "loss": 1.1227, | |
| "step": 1561 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.670466664858452e-07, | |
| "loss": 1.1959, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.61962184111625e-07, | |
| "loss": 1.1362, | |
| "step": 1563 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.5691251482424914e-07, | |
| "loss": 1.1523, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.5189767686338196e-07, | |
| "loss": 1.1422, | |
| "step": 1565 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.469176883428782e-07, | |
| "loss": 1.1937, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.419725672507068e-07, | |
| "loss": 1.1965, | |
| "step": 1567 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.3706233144890296e-07, | |
| "loss": 1.1909, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.3218699867348937e-07, | |
| "loss": 1.2193, | |
| "step": 1569 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.273465865344172e-07, | |
| "loss": 1.182, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.2254111251550424e-07, | |
| "loss": 1.1516, | |
| "step": 1571 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.1777059397436693e-07, | |
| "loss": 1.1567, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.1303504814236494e-07, | |
| "loss": 1.1832, | |
| "step": 1573 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.0833449212453237e-07, | |
| "loss": 1.16, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.0366894289951875e-07, | |
| "loss": 1.2157, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.99038417319526e-07, | |
| "loss": 1.0917, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 2.944429321102504e-07, | |
| "loss": 1.1507, | |
| "step": 1577 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.8988250387082083e-07, | |
| "loss": 1.1437, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.853571490737372e-07, | |
| "loss": 1.1051, | |
| "step": 1579 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.8086688406481546e-07, | |
| "loss": 1.123, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.764117250631182e-07, | |
| "loss": 1.2153, | |
| "step": 1581 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.7199168816091416e-07, | |
| "loss": 1.2143, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.6760678932359787e-07, | |
| "loss": 1.2127, | |
| "step": 1583 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.6325704438965206e-07, | |
| "loss": 1.205, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.589424690705777e-07, | |
| "loss": 1.1716, | |
| "step": 1585 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.5466307895084194e-07, | |
| "loss": 1.1912, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.5041888948782124e-07, | |
| "loss": 1.0927, | |
| "step": 1587 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.4620991601174596e-07, | |
| "loss": 1.1741, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.420361737256438e-07, | |
| "loss": 1.2194, | |
| "step": 1589 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.378976777052866e-07, | |
| "loss": 1.149, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.3379444289913344e-07, | |
| "loss": 1.1055, | |
| "step": 1591 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.2972648412827982e-07, | |
| "loss": 1.1654, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.256938160863986e-07, | |
| "loss": 1.2002, | |
| "step": 1593 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.2169645333969815e-07, | |
| "loss": 1.1869, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1773441032685417e-07, | |
| "loss": 1.2332, | |
| "step": 1595 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1380770135897122e-07, | |
| "loss": 1.1522, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0991634061952038e-07, | |
| "loss": 1.1533, | |
| "step": 1597 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.060603421643004e-07, | |
| "loss": 1.1986, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0223971992137325e-07, | |
| "loss": 1.1269, | |
| "step": 1599 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.9845448769102105e-07, | |
| "loss": 1.244, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.9470465914569915e-07, | |
| "loss": 1.2463, | |
| "step": 1601 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.9099024782997965e-07, | |
| "loss": 1.1368, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8731126716050707e-07, | |
| "loss": 1.1259, | |
| "step": 1603 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.836677304259471e-07, | |
| "loss": 1.1975, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.8005965078694233e-07, | |
| "loss": 1.2085, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.764870412760611e-07, | |
| "loss": 1.151, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7294991479775202e-07, | |
| "loss": 1.2416, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.694482841282974e-07, | |
| "loss": 1.0676, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.6598216191576643e-07, | |
| "loss": 1.2049, | |
| "step": 1609 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.6255156067997325e-07, | |
| "loss": 1.1505, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.5915649281242007e-07, | |
| "loss": 1.2177, | |
| "step": 1611 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5579697057627074e-07, | |
| "loss": 1.0982, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5247300610628957e-07, | |
| "loss": 1.1824, | |
| "step": 1613 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4918461140880691e-07, | |
| "loss": 1.019, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.459317983616737e-07, | |
| "loss": 1.166, | |
| "step": 1615 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.42714578714217e-07, | |
| "loss": 1.155, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3953296408720007e-07, | |
| "loss": 1.1519, | |
| "step": 1617 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3638696597277678e-07, | |
| "loss": 1.1228, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3327659573445395e-07, | |
| "loss": 1.1829, | |
| "step": 1619 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3020186460704908e-07, | |
| "loss": 1.1594, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2716278369664825e-07, | |
| "loss": 1.1898, | |
| "step": 1621 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2415936398056715e-07, | |
| "loss": 1.2258, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.2119161630731234e-07, | |
| "loss": 1.1862, | |
| "step": 1623 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.1825955139654121e-07, | |
| "loss": 1.2319, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.153631798390209e-07, | |
| "loss": 1.1628, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.1250251209659502e-07, | |
| "loss": 1.1546, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.0967755850214257e-07, | |
| "loss": 1.168, | |
| "step": 1627 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.0688832925954019e-07, | |
| "loss": 1.2396, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.041348344436277e-07, | |
| "loss": 1.1595, | |
| "step": 1629 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0141708400016714e-07, | |
| "loss": 1.116, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.873508774581597e-08, | |
| "loss": 1.1527, | |
| "step": 1631 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.608885536808055e-08, | |
| "loss": 1.1513, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.347839642528721e-08, | |
| "loss": 1.1576, | |
| "step": 1633 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.090372034654793e-08, | |
| "loss": 1.0495, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.836483643172577e-08, | |
| "loss": 1.1289, | |
| "step": 1635 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.58617538513995e-08, | |
| "loss": 1.0873, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.339448164682906e-08, | |
| "loss": 1.1691, | |
| "step": 1637 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.096302872993234e-08, | |
| "loss": 1.1659, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.856740388323847e-08, | |
| "loss": 1.18, | |
| "step": 1639 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.620761575987013e-08, | |
| "loss": 1.2707, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.388367288350018e-08, | |
| "loss": 1.2355, | |
| "step": 1641 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 7.159558364833175e-08, | |
| "loss": 1.2193, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 6.934335631905931e-08, | |
| "loss": 1.1395, | |
| "step": 1643 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 6.71269990308443e-08, | |
| "loss": 1.2028, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 6.494651978928179e-08, | |
| "loss": 1.1534, | |
| "step": 1645 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.280192647037498e-08, | |
| "loss": 1.1932, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.069322682050516e-08, | |
| "loss": 1.1857, | |
| "step": 1647 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.862042845640403e-08, | |
| "loss": 1.1041, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.65835388651248e-08, | |
| "loss": 1.1861, | |
| "step": 1649 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.458256540401774e-08, | |
| "loss": 1.1347, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.261751530070136e-08, | |
| "loss": 1.1571, | |
| "step": 1651 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.0688395653036846e-08, | |
| "loss": 1.1796, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.879521342910476e-08, | |
| "loss": 1.1338, | |
| "step": 1653 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.693797546717505e-08, | |
| "loss": 1.1761, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.5116688475685954e-08, | |
| "loss": 1.2456, | |
| "step": 1655 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.333135903322072e-08, | |
| "loss": 1.2409, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.1581993588482025e-08, | |
| "loss": 1.131, | |
| "step": 1657 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.986859846026425e-08, | |
| "loss": 1.2769, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.819117983744125e-08, | |
| "loss": 1.1935, | |
| "step": 1659 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.6549743778934166e-08, | |
| "loss": 1.1127, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.494429621369366e-08, | |
| "loss": 1.1966, | |
| "step": 1661 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.3374842940678834e-08, | |
| "loss": 1.1531, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.18413896288372e-08, | |
| "loss": 1.1041, | |
| "step": 1663 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.034394181707812e-08, | |
| "loss": 1.1673, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.888250491426159e-08, | |
| "loss": 1.1569, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.74570841991717e-08, | |
| "loss": 1.1125, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.606768482050215e-08, | |
| "loss": 1.1862, | |
| "step": 1667 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.4714311796834035e-08, | |
| "loss": 1.1537, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.3396970016619225e-08, | |
| "loss": 1.2256, | |
| "step": 1669 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.211566423816591e-08, | |
| "loss": 1.1421, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.08703990896153e-08, | |
| "loss": 1.1489, | |
| "step": 1671 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.9661179068928283e-08, | |
| "loss": 1.2294, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.8488008543871007e-08, | |
| "loss": 1.0836, | |
| "step": 1673 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7350891751993782e-08, | |
| "loss": 1.2013, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.624983280062331e-08, | |
| "loss": 1.3023, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.518483566683826e-08, | |
| "loss": 1.1781, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.415590419746371e-08, | |
| "loss": 1.2011, | |
| "step": 1677 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.3163042109053393e-08, | |
| "loss": 1.2066, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2206252987876366e-08, | |
| "loss": 1.1358, | |
| "step": 1679 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.1285540289903696e-08, | |
| "loss": 1.0472, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.040090734079624e-08, | |
| "loss": 1.1751, | |
| "step": 1681 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.552357335893547e-09, | |
| "loss": 1.1809, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.739893340202754e-09, | |
| "loss": 1.1425, | |
| "step": 1683 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.963518288385264e-09, | |
| "loss": 1.2003, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.223234984747862e-09, | |
| "loss": 1.1713, | |
| "step": 1685 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.5190461032305085e-09, | |
| "loss": 1.1553, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.850954187399671e-09, | |
| "loss": 1.1377, | |
| "step": 1687 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.218961650439447e-09, | |
| "loss": 1.1135, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.6230707751371284e-09, | |
| "loss": 1.1009, | |
| "step": 1689 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.063283713883203e-09, | |
| "loss": 1.2446, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.539602488654703e-09, | |
| "loss": 1.1807, | |
| "step": 1691 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.0520289910174194e-09, | |
| "loss": 1.2118, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.600564982110365e-09, | |
| "loss": 1.1417, | |
| "step": 1693 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.185212092645772e-09, | |
| "loss": 1.2131, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.8059718228979894e-09, | |
| "loss": 1.182, | |
| "step": 1695 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.462845542704594e-09, | |
| "loss": 1.0384, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.1558344914530673e-09, | |
| "loss": 1.1252, | |
| "step": 1697 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.849397780841262e-10, | |
| "loss": 1.2233, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 6.50162381083952e-10, | |
| "loss": 1.1729, | |
| "step": 1699 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.5150314847863806e-10, | |
| "loss": 1.2093, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.8896279783863224e-10, | |
| "loss": 1.0945, | |
| "step": 1701 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.6254191626430272e-10, | |
| "loss": 1.2, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 7.224096039704088e-11, | |
| "loss": 1.1717, | |
| "step": 1703 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.8060256408158892e-11, | |
| "loss": 1.251, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0, | |
| "loss": 1.1397, | |
| "step": 1705 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 1705, | |
| "total_flos": 307764793589760.0, | |
| "train_loss": 1.249325556419462, | |
| "train_runtime": 24639.388, | |
| "train_samples_per_second": 16.599, | |
| "train_steps_per_second": 0.069 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 1705, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 50000, | |
| "total_flos": 307764793589760.0, | |
| "train_batch_size": 24, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |