| { | |
| "best_metric": 0.98, | |
| "best_model_checkpoint": "timesformer-base-finetuned-k400-finetuned-kinetic400-epoch6-num_frame_10_TimeSformer_more_data/checkpoint-1250", | |
| "epoch": 5.166666666666667, | |
| "eval_steps": 500, | |
| "global_step": 1500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0, | |
| "step": 0 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 1 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 2 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 3 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 4 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 5 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 6 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.0, | |
| "step": 7 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 8 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.01, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "grad_norm": 23.523942947387695, | |
| "learning_rate": 3.3333333333333333e-06, | |
| "loss": 1.9753, | |
| "step": 10 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 10 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 11 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 12 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 13 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 14 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 15 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 16 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 17 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 18 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "grad_norm": 31.144126892089844, | |
| "learning_rate": 6.666666666666667e-06, | |
| "loss": 2.017, | |
| "step": 20 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 20 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.01, | |
| "step": 21 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.01, | |
| "step": 22 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 23 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.02, | |
| "step": 24 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 25 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 26 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 27 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 28 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "grad_norm": 29.392616271972656, | |
| "learning_rate": 1e-05, | |
| "loss": 1.5881, | |
| "step": 30 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 30 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.02, | |
| "step": 31 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 32 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.02, | |
| "step": 33 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.02, | |
| "step": 34 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.02, | |
| "step": 35 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 36 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.02, | |
| "step": 37 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.03, | |
| "step": 38 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "grad_norm": 28.781667709350586, | |
| "learning_rate": 1.3333333333333333e-05, | |
| "loss": 1.4893, | |
| "step": 40 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 40 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 41 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 42 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.03, | |
| "step": 43 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 44 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 45 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 46 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 47 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 48 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "grad_norm": 16.677244186401367, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 1.0796, | |
| "step": 50 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.03, | |
| "step": 50 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.03, | |
| "step": 51 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.03, | |
| "step": 52 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.04, | |
| "step": 53 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 54 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 55 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 56 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 57 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 58 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.04, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "grad_norm": 31.004791259765625, | |
| "learning_rate": 2e-05, | |
| "loss": 1.0754, | |
| "step": 60 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 60 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 61 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 62 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.04, | |
| "step": 63 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 64 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 65 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.04, | |
| "step": 66 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.04, | |
| "step": 67 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 68 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "grad_norm": 8.404767036437988, | |
| "learning_rate": 2.3333333333333336e-05, | |
| "loss": 0.8469, | |
| "step": 70 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 70 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 71 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 72 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.05, | |
| "step": 73 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 74 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 75 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 76 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 77 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 78 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "grad_norm": 0.5016779899597168, | |
| "learning_rate": 2.6666666666666667e-05, | |
| "loss": 0.3259, | |
| "step": 80 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 80 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 81 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.05, | |
| "step": 82 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 83 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 84 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 85 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 86 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 87 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 88 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "grad_norm": 1.0599085092544556, | |
| "learning_rate": 3e-05, | |
| "loss": 0.2104, | |
| "step": 90 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 90 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 91 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 92 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.06, | |
| "step": 93 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 94 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 95 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 96 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.06, | |
| "step": 97 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 98 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "grad_norm": 0.25412893295288086, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.2345, | |
| "step": 100 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 100 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 101 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 102 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 103 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 104 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 105 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.07, | |
| "step": 106 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 107 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 108 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "grad_norm": 33.525970458984375, | |
| "learning_rate": 3.6666666666666666e-05, | |
| "loss": 0.1531, | |
| "step": 110 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 110 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.07, | |
| "step": 111 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.07, | |
| "step": 112 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 113 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 114 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 115 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 116 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 117 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 118 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "grad_norm": 1.2523601055145264, | |
| "learning_rate": 4e-05, | |
| "loss": 0.3343, | |
| "step": 120 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 120 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 121 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 122 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 123 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 124 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 125 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 126 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.08, | |
| "step": 127 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 128 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "grad_norm": 0.017991458997130394, | |
| "learning_rate": 4.3333333333333334e-05, | |
| "loss": 0.1481, | |
| "step": 130 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 130 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 131 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.09, | |
| "step": 132 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 133 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 134 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 135 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 136 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 137 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 138 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.09, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "grad_norm": 63.53673553466797, | |
| "learning_rate": 4.666666666666667e-05, | |
| "loss": 0.4044, | |
| "step": 140 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.09, | |
| "step": 140 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 141 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.09, | |
| "step": 142 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 143 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.1, | |
| "step": 144 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 145 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 146 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.1, | |
| "step": 147 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 148 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "grad_norm": 0.6711806058883667, | |
| "learning_rate": 5e-05, | |
| "loss": 1.0807, | |
| "step": 150 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 150 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.1, | |
| "step": 151 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 152 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 153 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 154 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 155 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.1, | |
| "step": 156 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.1, | |
| "step": 157 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 158 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "grad_norm": 1.7794537544250488, | |
| "learning_rate": 4.962962962962963e-05, | |
| "loss": 0.4639, | |
| "step": 160 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 160 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 161 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 162 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 163 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.11, | |
| "step": 164 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 165 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 166 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 167 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.11, | |
| "step": 168 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.11, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "grad_norm": 71.7469711303711, | |
| "learning_rate": 4.925925925925926e-05, | |
| "loss": 1.0374, | |
| "step": 170 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 170 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 171 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.11, | |
| "step": 172 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 173 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 174 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 175 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 176 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 177 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 178 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "grad_norm": 0.37160196900367737, | |
| "learning_rate": 4.888888888888889e-05, | |
| "loss": 0.0049, | |
| "step": 180 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 180 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 181 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 182 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 183 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 184 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 185 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 186 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.12, | |
| "step": 187 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 188 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "grad_norm": 0.050026535987854004, | |
| "learning_rate": 4.851851851851852e-05, | |
| "loss": 0.054, | |
| "step": 190 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 190 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 191 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 192 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 193 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 194 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 195 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 196 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 197 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 198 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "grad_norm": 0.02174362912774086, | |
| "learning_rate": 4.814814814814815e-05, | |
| "loss": 0.0625, | |
| "step": 200 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.13, | |
| "step": 200 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 201 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.13, | |
| "step": 202 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 203 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 204 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 205 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 206 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 207 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 208 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "grad_norm": 0.03173552453517914, | |
| "learning_rate": 4.7777777777777784e-05, | |
| "loss": 0.3317, | |
| "step": 210 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 210 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 211 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 212 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.14, | |
| "step": 213 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 214 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 215 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 216 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.14, | |
| "step": 217 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 218 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "grad_norm": 0.7077636122703552, | |
| "learning_rate": 4.740740740740741e-05, | |
| "loss": 0.4387, | |
| "step": 220 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.15, | |
| "step": 220 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 221 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 222 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 223 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 224 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 225 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 226 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 227 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 228 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "grad_norm": 0.05836891755461693, | |
| "learning_rate": 4.703703703703704e-05, | |
| "loss": 0.6192, | |
| "step": 230 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 230 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 231 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.15, | |
| "step": 232 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 233 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 234 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 235 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 236 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 237 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 238 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "grad_norm": 0.07609537243843079, | |
| "learning_rate": 4.666666666666667e-05, | |
| "loss": 0.0226, | |
| "step": 240 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 240 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 241 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 242 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 243 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.16, | |
| "step": 244 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 245 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 246 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.16, | |
| "step": 247 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 248 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "grad_norm": 0.1481229066848755, | |
| "learning_rate": 4.62962962962963e-05, | |
| "loss": 0.3263, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_accuracy": 0.95, | |
| "eval_loss": 0.1772526353597641, | |
| "eval_runtime": 106.8573, | |
| "eval_samples_per_second": 0.936, | |
| "eval_steps_per_second": 0.936, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 0.17, | |
| "step": 250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.0, | |
| "step": 251 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.0, | |
| "step": 252 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.0, | |
| "step": 253 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.0, | |
| "step": 254 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.0, | |
| "step": 255 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.0, | |
| "step": 256 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.0, | |
| "step": 257 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 258 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "grad_norm": 0.4767157733440399, | |
| "learning_rate": 4.592592592592593e-05, | |
| "loss": 0.3562, | |
| "step": 260 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 260 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 261 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 262 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.01, | |
| "step": 263 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 264 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 265 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 266 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 267 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 268 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "grad_norm": 0.012715524062514305, | |
| "learning_rate": 4.555555555555556e-05, | |
| "loss": 0.68, | |
| "step": 270 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 270 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 271 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.01, | |
| "step": 272 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 273 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 274 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 275 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 276 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 277 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 278 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "grad_norm": 0.006971323862671852, | |
| "learning_rate": 4.518518518518519e-05, | |
| "loss": 0.0035, | |
| "step": 280 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 280 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 281 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 282 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 283 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 284 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 285 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 286 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.02, | |
| "step": 287 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 288 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "grad_norm": 0.005007774569094181, | |
| "learning_rate": 4.481481481481482e-05, | |
| "loss": 0.0004, | |
| "step": 290 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 290 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 291 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 292 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 293 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 294 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 295 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 296 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 297 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 298 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "grad_norm": 0.00695458147674799, | |
| "learning_rate": 4.4444444444444447e-05, | |
| "loss": 0.0172, | |
| "step": 300 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 300 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 301 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.03, | |
| "step": 302 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 303 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 304 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 305 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 306 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 307 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 308 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "grad_norm": 0.019786611199378967, | |
| "learning_rate": 4.4074074074074076e-05, | |
| "loss": 0.0005, | |
| "step": 310 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 310 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 311 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 312 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 313 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 314 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 315 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 316 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.04, | |
| "step": 317 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 318 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "grad_norm": 0.008017357438802719, | |
| "learning_rate": 4.3703703703703705e-05, | |
| "loss": 0.0615, | |
| "step": 320 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 320 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 321 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 322 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 323 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 324 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 325 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 326 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 327 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 328 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "grad_norm": 0.00832881685346365, | |
| "learning_rate": 4.3333333333333334e-05, | |
| "loss": 0.0006, | |
| "step": 330 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 330 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 331 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.05, | |
| "step": 332 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 333 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 334 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.06, | |
| "step": 335 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 336 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 337 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 338 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "grad_norm": 0.006028041709214449, | |
| "learning_rate": 4.296296296296296e-05, | |
| "loss": 0.1193, | |
| "step": 340 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 340 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 341 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.06, | |
| "step": 342 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 343 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 344 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 345 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 346 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.06, | |
| "step": 347 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 348 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "grad_norm": 0.09666573256254196, | |
| "learning_rate": 4.259259259259259e-05, | |
| "loss": 0.1317, | |
| "step": 350 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 350 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 351 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 352 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 353 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 354 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 355 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 356 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 357 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 358 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "grad_norm": 0.006060417275875807, | |
| "learning_rate": 4.222222222222222e-05, | |
| "loss": 0.0055, | |
| "step": 360 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 360 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 361 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.07, | |
| "step": 362 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 363 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 364 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 365 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 366 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 367 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 368 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "grad_norm": 0.0059410142712295055, | |
| "learning_rate": 4.185185185185185e-05, | |
| "loss": 0.0257, | |
| "step": 370 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 370 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 371 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 372 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 373 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 374 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 375 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 376 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.08, | |
| "step": 377 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 378 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "grad_norm": 0.009067313745617867, | |
| "learning_rate": 4.148148148148148e-05, | |
| "loss": 0.009, | |
| "step": 380 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 380 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 381 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 382 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 383 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 384 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 385 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 386 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 387 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 388 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "grad_norm": 0.0041416967287659645, | |
| "learning_rate": 4.111111111111111e-05, | |
| "loss": 0.0073, | |
| "step": 390 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 390 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 391 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.09, | |
| "step": 392 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 393 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 394 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.1, | |
| "step": 395 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 396 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 397 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 398 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "grad_norm": 0.01870698295533657, | |
| "learning_rate": 4.074074074074074e-05, | |
| "loss": 0.6573, | |
| "step": 400 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 400 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.1, | |
| "step": 401 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 402 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 403 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 404 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 405 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 406 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.1, | |
| "step": 407 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 408 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.11, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "grad_norm": 113.8437271118164, | |
| "learning_rate": 4.0370370370370374e-05, | |
| "loss": 0.9539, | |
| "step": 410 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 410 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 411 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 412 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 413 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.11, | |
| "step": 414 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 415 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.11, | |
| "step": 416 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 417 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 418 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "grad_norm": 0.018104052171111107, | |
| "learning_rate": 4e-05, | |
| "loss": 0.5387, | |
| "step": 420 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 420 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 421 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.11, | |
| "step": 422 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 423 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 424 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 425 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 426 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 427 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 428 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "grad_norm": 0.003394722007215023, | |
| "learning_rate": 3.962962962962963e-05, | |
| "loss": 0.0007, | |
| "step": 430 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 430 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 431 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 432 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 433 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 434 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 435 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 436 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.12, | |
| "step": 437 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 438 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "grad_norm": 0.013579904101788998, | |
| "learning_rate": 3.925925925925926e-05, | |
| "loss": 0.0011, | |
| "step": 440 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 440 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 441 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 442 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 443 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 444 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 445 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 446 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 447 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 448 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "grad_norm": 120.73057556152344, | |
| "learning_rate": 3.888888888888889e-05, | |
| "loss": 0.0674, | |
| "step": 450 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 450 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 451 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.13, | |
| "step": 452 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 453 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 454 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 455 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 456 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 457 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 458 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "grad_norm": 0.03209913149476051, | |
| "learning_rate": 3.851851851851852e-05, | |
| "loss": 0.0022, | |
| "step": 460 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 460 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 461 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 462 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 463 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 464 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 465 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 466 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.14, | |
| "step": 467 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 468 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "grad_norm": 0.004453280940651894, | |
| "learning_rate": 3.814814814814815e-05, | |
| "loss": 0.0158, | |
| "step": 470 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 470 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 471 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 472 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 473 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 474 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 475 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 476 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 477 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 478 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "grad_norm": 0.013010458089411259, | |
| "learning_rate": 3.777777777777778e-05, | |
| "loss": 0.0112, | |
| "step": 480 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 480 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 481 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.15, | |
| "step": 482 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 483 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 484 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 485 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 486 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 487 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 488 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "grad_norm": 0.21454155445098877, | |
| "learning_rate": 3.740740740740741e-05, | |
| "loss": 0.0648, | |
| "step": 490 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 490 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 491 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.16, | |
| "step": 492 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 493 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 494 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 495 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 496 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.16, | |
| "step": 497 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 498 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "grad_norm": 0.0049698748625814915, | |
| "learning_rate": 3.7037037037037037e-05, | |
| "loss": 0.367, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "eval_accuracy": 0.91, | |
| "eval_loss": 0.611099898815155, | |
| "eval_runtime": 107.7031, | |
| "eval_samples_per_second": 0.928, | |
| "eval_steps_per_second": 0.928, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 1.17, | |
| "step": 500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 501 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 502 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 503 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 504 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 505 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 506 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.0, | |
| "step": 507 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 508 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "grad_norm": 0.011625483632087708, | |
| "learning_rate": 3.6666666666666666e-05, | |
| "loss": 0.0165, | |
| "step": 510 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 510 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 511 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 512 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 513 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 514 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 515 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 516 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 517 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 518 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "grad_norm": 0.0068740141578018665, | |
| "learning_rate": 3.62962962962963e-05, | |
| "loss": 0.0002, | |
| "step": 520 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 520 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 521 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.01, | |
| "step": 522 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 523 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 524 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 525 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 526 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 527 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 528 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "grad_norm": 0.002710917266085744, | |
| "learning_rate": 3.592592592592593e-05, | |
| "loss": 0.0058, | |
| "step": 530 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 530 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 531 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 532 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 533 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 534 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 535 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 536 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.02, | |
| "step": 537 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 538 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "grad_norm": 0.014188497327268124, | |
| "learning_rate": 3.555555555555556e-05, | |
| "loss": 0.0033, | |
| "step": 540 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 540 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 541 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 542 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 543 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 544 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 545 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 546 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 547 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 548 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "grad_norm": 0.005379448179155588, | |
| "learning_rate": 3.518518518518519e-05, | |
| "loss": 0.0026, | |
| "step": 550 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 550 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.03, | |
| "step": 551 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.03, | |
| "step": 552 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 553 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 554 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 555 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 556 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 557 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 558 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "grad_norm": 37.30723190307617, | |
| "learning_rate": 3.481481481481482e-05, | |
| "loss": 0.34, | |
| "step": 560 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 560 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 561 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 562 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 563 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 564 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 565 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 566 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.04, | |
| "step": 567 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 568 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "grad_norm": 0.005344702396541834, | |
| "learning_rate": 3.444444444444445e-05, | |
| "loss": 0.0188, | |
| "step": 570 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 570 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 571 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 572 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 573 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 574 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 575 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 576 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 577 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 578 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "grad_norm": 0.002420774195343256, | |
| "learning_rate": 3.4074074074074077e-05, | |
| "loss": 0.0016, | |
| "step": 580 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 580 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 581 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.05, | |
| "step": 582 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 583 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 584 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 585 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 586 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 587 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 588 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "grad_norm": 0.004318414255976677, | |
| "learning_rate": 3.3703703703703706e-05, | |
| "loss": 0.0003, | |
| "step": 590 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 590 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 591 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 592 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 593 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 594 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 595 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 596 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.06, | |
| "step": 597 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 598 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "grad_norm": 0.008985906839370728, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.0007, | |
| "step": 600 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 600 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 601 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 602 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 603 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 604 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 605 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 606 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 607 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 608 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "grad_norm": 0.004877213388681412, | |
| "learning_rate": 3.2962962962962964e-05, | |
| "loss": 0.0005, | |
| "step": 610 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 610 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 611 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.07, | |
| "step": 612 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 613 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 614 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 615 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 616 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 617 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 618 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "grad_norm": 0.009020144119858742, | |
| "learning_rate": 3.25925925925926e-05, | |
| "loss": 0.0009, | |
| "step": 620 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 620 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 621 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 622 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.08, | |
| "step": 623 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 624 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 625 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 626 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.08, | |
| "step": 627 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 628 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "grad_norm": 0.004489729646593332, | |
| "learning_rate": 3.222222222222223e-05, | |
| "loss": 0.5665, | |
| "step": 630 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 630 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 631 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 632 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.09, | |
| "step": 633 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 634 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 635 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 636 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 637 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 638 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "grad_norm": 0.0066104307770729065, | |
| "learning_rate": 3.185185185185185e-05, | |
| "loss": 0.0838, | |
| "step": 640 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 640 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 641 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.09, | |
| "step": 642 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 643 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 644 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 645 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 646 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 647 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 648 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "grad_norm": 0.002852922771126032, | |
| "learning_rate": 3.148148148148148e-05, | |
| "loss": 0.0001, | |
| "step": 650 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 650 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 651 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 652 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 653 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 654 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 655 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 656 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.1, | |
| "step": 657 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 658 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "grad_norm": 0.006365996785461903, | |
| "learning_rate": 3.111111111111111e-05, | |
| "loss": 0.0003, | |
| "step": 660 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 660 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 661 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 662 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 663 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 664 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.11, | |
| "step": 665 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 666 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 667 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 668 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "grad_norm": 0.0015251237200573087, | |
| "learning_rate": 3.074074074074074e-05, | |
| "loss": 0.2052, | |
| "step": 670 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 670 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 671 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.11, | |
| "step": 672 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 673 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 674 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 675 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 676 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 677 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 678 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "grad_norm": 0.011465414427220821, | |
| "learning_rate": 3.037037037037037e-05, | |
| "loss": 0.0017, | |
| "step": 680 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 680 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 681 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 682 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 683 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 684 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 685 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 686 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.12, | |
| "step": 687 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 688 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "grad_norm": 0.001982511952519417, | |
| "learning_rate": 3e-05, | |
| "loss": 0.0001, | |
| "step": 690 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 690 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 691 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 692 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 693 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 694 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 695 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 696 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 697 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 698 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "grad_norm": 0.0018550872337073088, | |
| "learning_rate": 2.962962962962963e-05, | |
| "loss": 0.0006, | |
| "step": 700 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 700 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 701 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.13, | |
| "step": 702 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 703 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 704 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 705 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 706 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 707 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 708 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "grad_norm": 0.009017335250973701, | |
| "learning_rate": 2.925925925925926e-05, | |
| "loss": 0.0001, | |
| "step": 710 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 710 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 711 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 712 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 713 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 714 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 715 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 716 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.14, | |
| "step": 717 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 718 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "grad_norm": 0.0024712635204195976, | |
| "learning_rate": 2.8888888888888888e-05, | |
| "loss": 0.0002, | |
| "step": 720 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 720 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 721 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 722 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 723 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 724 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 725 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 726 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 727 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 728 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "grad_norm": 0.004336017183959484, | |
| "learning_rate": 2.851851851851852e-05, | |
| "loss": 0.0007, | |
| "step": 730 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 730 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 731 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.15, | |
| "step": 732 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 733 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 734 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 735 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 736 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 737 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 738 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "grad_norm": 0.0031830829102545977, | |
| "learning_rate": 2.814814814814815e-05, | |
| "loss": 0.0458, | |
| "step": 740 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 740 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 741 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 742 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 743 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 744 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 745 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 746 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.16, | |
| "step": 747 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 748 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "grad_norm": 0.0018775537610054016, | |
| "learning_rate": 2.777777777777778e-05, | |
| "loss": 0.0003, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "eval_accuracy": 0.95, | |
| "eval_loss": 0.3224475383758545, | |
| "eval_runtime": 106.1095, | |
| "eval_samples_per_second": 0.942, | |
| "eval_steps_per_second": 0.942, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 2.17, | |
| "step": 750 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 751 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 752 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 753 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 754 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 755 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 756 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.0, | |
| "step": 757 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 758 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "grad_norm": 0.0055166068486869335, | |
| "learning_rate": 2.7407407407407408e-05, | |
| "loss": 0.0001, | |
| "step": 760 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.01, | |
| "step": 760 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 761 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 762 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 763 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 764 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 765 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 766 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 767 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 768 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "grad_norm": 0.019787773489952087, | |
| "learning_rate": 2.7037037037037037e-05, | |
| "loss": 0.6311, | |
| "step": 770 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 770 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 771 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.01, | |
| "step": 772 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.02, | |
| "step": 773 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 774 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 775 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 776 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 777 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 778 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "grad_norm": 0.008881332352757454, | |
| "learning_rate": 2.6666666666666667e-05, | |
| "loss": 0.0912, | |
| "step": 780 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 780 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 781 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.02, | |
| "step": 782 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 783 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 784 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 785 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 786 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.02, | |
| "step": 787 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 788 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "grad_norm": 0.0062670777551829815, | |
| "learning_rate": 2.6296296296296296e-05, | |
| "loss": 0.597, | |
| "step": 790 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 790 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 791 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 792 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 793 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 794 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 795 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 796 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 797 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 798 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "grad_norm": 0.008829276077449322, | |
| "learning_rate": 2.5925925925925925e-05, | |
| "loss": 0.0006, | |
| "step": 800 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 800 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 801 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.03, | |
| "step": 802 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 803 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 804 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 805 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 806 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 807 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 808 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "grad_norm": 0.004734104033559561, | |
| "learning_rate": 2.5555555555555554e-05, | |
| "loss": 0.0001, | |
| "step": 810 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 810 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 811 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.04, | |
| "step": 812 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 813 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 814 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 815 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 816 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.04, | |
| "step": 817 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 818 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "grad_norm": 0.0033767791464924812, | |
| "learning_rate": 2.5185185185185183e-05, | |
| "loss": 0.3228, | |
| "step": 820 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 820 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 821 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 822 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 823 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 824 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 825 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 826 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 827 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 828 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "grad_norm": 0.006589946802705526, | |
| "learning_rate": 2.4814814814814816e-05, | |
| "loss": 0.0008, | |
| "step": 830 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 830 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 831 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.05, | |
| "step": 832 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 833 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 834 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 835 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 836 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 837 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 838 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "grad_norm": 0.034783441573381424, | |
| "learning_rate": 2.4444444444444445e-05, | |
| "loss": 0.0007, | |
| "step": 840 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 840 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 841 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 842 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 843 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 844 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 845 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 846 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.06, | |
| "step": 847 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 848 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "grad_norm": 0.13152948021888733, | |
| "learning_rate": 2.4074074074074074e-05, | |
| "loss": 0.0004, | |
| "step": 850 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 850 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 851 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 852 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 853 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 854 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 855 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 856 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 857 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 858 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "grad_norm": 0.0013600288657471538, | |
| "learning_rate": 2.3703703703703707e-05, | |
| "loss": 0.0002, | |
| "step": 860 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 860 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 861 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.07, | |
| "step": 862 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 863 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 864 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 865 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 866 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 867 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 868 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "grad_norm": 0.0017450809245929122, | |
| "learning_rate": 2.3333333333333336e-05, | |
| "loss": 0.0001, | |
| "step": 870 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 870 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 871 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 872 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 873 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 874 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.08, | |
| "step": 875 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 876 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.08, | |
| "step": 877 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 878 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "grad_norm": 0.0039650010876357555, | |
| "learning_rate": 2.2962962962962965e-05, | |
| "loss": 0.4801, | |
| "step": 880 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 880 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 881 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 882 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 883 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 884 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 885 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 886 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 887 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 888 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "grad_norm": 0.0033429365139454603, | |
| "learning_rate": 2.2592592592592594e-05, | |
| "loss": 0.0001, | |
| "step": 890 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 890 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 891 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.09, | |
| "step": 892 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 893 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 894 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 895 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 896 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 897 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 898 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "grad_norm": 0.02160213515162468, | |
| "learning_rate": 2.2222222222222223e-05, | |
| "loss": 0.0001, | |
| "step": 900 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 900 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 901 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 902 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 903 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 904 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 905 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 906 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.1, | |
| "step": 907 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 908 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "grad_norm": 0.03846479207277298, | |
| "learning_rate": 2.1851851851851852e-05, | |
| "loss": 0.0002, | |
| "step": 910 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 910 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 911 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 912 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 913 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 914 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 915 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 916 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 917 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 918 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "grad_norm": 0.0017593818483874202, | |
| "learning_rate": 2.148148148148148e-05, | |
| "loss": 0.0002, | |
| "step": 920 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 920 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 921 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.11, | |
| "step": 922 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 923 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 924 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 925 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 926 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 927 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 928 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "grad_norm": 0.0029362498316913843, | |
| "learning_rate": 2.111111111111111e-05, | |
| "loss": 0.0002, | |
| "step": 930 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 930 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 931 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 932 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 933 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 934 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 935 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 936 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.12, | |
| "step": 937 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 938 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "grad_norm": 0.0026593152433633804, | |
| "learning_rate": 2.074074074074074e-05, | |
| "loss": 0.0005, | |
| "step": 940 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 940 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 941 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 942 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 943 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 944 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 945 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 946 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 947 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 948 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "grad_norm": 0.02160882204771042, | |
| "learning_rate": 2.037037037037037e-05, | |
| "loss": 0.0005, | |
| "step": 950 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 950 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 951 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.13, | |
| "step": 952 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 953 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 954 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 955 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 956 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 957 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 958 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "grad_norm": 0.09423842281103134, | |
| "learning_rate": 2e-05, | |
| "loss": 0.0003, | |
| "step": 960 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 960 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 961 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 962 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 963 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 964 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 965 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 966 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.14, | |
| "step": 967 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 968 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "grad_norm": 0.0020063407719135284, | |
| "learning_rate": 1.962962962962963e-05, | |
| "loss": 0.0001, | |
| "step": 970 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 970 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 971 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 972 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 973 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 974 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 975 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 976 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 977 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 978 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "grad_norm": 0.02009579725563526, | |
| "learning_rate": 1.925925925925926e-05, | |
| "loss": 0.0009, | |
| "step": 980 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 980 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 981 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.15, | |
| "step": 982 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 983 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 984 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 985 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 986 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 987 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 988 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "grad_norm": 0.0024299276992678642, | |
| "learning_rate": 1.888888888888889e-05, | |
| "loss": 0.0003, | |
| "step": 990 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 990 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 991 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 992 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 993 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 994 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 995 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 996 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.16, | |
| "step": 997 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 998 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "grad_norm": 0.002172068227082491, | |
| "learning_rate": 1.8518518518518518e-05, | |
| "loss": 0.0006, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "eval_accuracy": 0.96, | |
| "eval_loss": 0.20253026485443115, | |
| "eval_runtime": 108.4279, | |
| "eval_samples_per_second": 0.922, | |
| "eval_steps_per_second": 0.922, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 3.17, | |
| "step": 1000 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1001 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1002 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1003 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1004 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1005 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1006 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.0, | |
| "step": 1007 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1008 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "grad_norm": 0.00461992435157299, | |
| "learning_rate": 1.814814814814815e-05, | |
| "loss": 0.0001, | |
| "step": 1010 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1010 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1011 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1012 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1013 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1014 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1015 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1016 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1017 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1018 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "grad_norm": 0.0016442248597741127, | |
| "learning_rate": 1.777777777777778e-05, | |
| "loss": 0.0001, | |
| "step": 1020 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1020 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1021 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.01, | |
| "step": 1022 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1023 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1024 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1025 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1026 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1027 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1028 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "grad_norm": 0.003995942417532206, | |
| "learning_rate": 1.740740740740741e-05, | |
| "loss": 0.0001, | |
| "step": 1030 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1030 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1031 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1032 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1033 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1034 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1035 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1036 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.02, | |
| "step": 1037 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1038 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "grad_norm": 0.0013397905277088284, | |
| "learning_rate": 1.7037037037037038e-05, | |
| "loss": 0.0001, | |
| "step": 1040 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1040 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1041 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1042 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1043 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1044 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1045 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1046 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1047 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1048 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "grad_norm": 0.07875602692365646, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 0.0002, | |
| "step": 1050 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1050 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1051 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.03, | |
| "step": 1052 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1053 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1054 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1055 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1056 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1057 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1058 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "grad_norm": 0.0040701306425035, | |
| "learning_rate": 1.62962962962963e-05, | |
| "loss": 0.0001, | |
| "step": 1060 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1060 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1061 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1062 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1063 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1064 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1065 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1066 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.04, | |
| "step": 1067 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1068 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "grad_norm": 0.03816696256399155, | |
| "learning_rate": 1.5925925925925926e-05, | |
| "loss": 0.0002, | |
| "step": 1070 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1070 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1071 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1072 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1073 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1074 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1075 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1076 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1077 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1078 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "grad_norm": 0.0012417088728398085, | |
| "learning_rate": 1.5555555555555555e-05, | |
| "loss": 0.0002, | |
| "step": 1080 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1080 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1081 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.05, | |
| "step": 1082 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1083 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1084 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1085 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1086 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1087 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1088 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 4.06, | |
| "grad_norm": 0.0019886523950845003, | |
| "learning_rate": 1.5185185185185186e-05, | |
| "loss": 0.0001, | |
| "step": 1090 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1090 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1091 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1092 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1093 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1094 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1095 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1096 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.06, | |
| "step": 1097 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1098 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "grad_norm": 0.012906285934150219, | |
| "learning_rate": 1.4814814814814815e-05, | |
| "loss": 0.0001, | |
| "step": 1100 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1100 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1101 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1102 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1103 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1104 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1105 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1106 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1107 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1108 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 4.07, | |
| "grad_norm": 0.0028517385944724083, | |
| "learning_rate": 1.4444444444444444e-05, | |
| "loss": 0.0064, | |
| "step": 1110 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1110 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1111 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.07, | |
| "step": 1112 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1113 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1114 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1115 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1116 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1117 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1118 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "grad_norm": 0.00149582140147686, | |
| "learning_rate": 1.4074074074074075e-05, | |
| "loss": 0.0001, | |
| "step": 1120 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1120 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1121 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1122 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1123 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1124 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1125 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1126 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.08, | |
| "step": 1127 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1128 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "grad_norm": 0.0013316815020516515, | |
| "learning_rate": 1.3703703703703704e-05, | |
| "loss": 0.0001, | |
| "step": 1130 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1130 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1131 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1132 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1133 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1134 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1135 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1136 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1137 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1138 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 4.09, | |
| "grad_norm": 0.002085028914734721, | |
| "learning_rate": 1.3333333333333333e-05, | |
| "loss": 0.0001, | |
| "step": 1140 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1140 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 4.09, | |
| "step": 1141 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.09, | |
| "step": 1142 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1143 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1144 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1145 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1146 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 4.1, | |
| "step": 1147 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1148 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "grad_norm": 0.004692728631198406, | |
| "learning_rate": 1.2962962962962962e-05, | |
| "loss": 0.5534, | |
| "step": 1150 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1150 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1151 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1152 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1153 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1154 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1155 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1156 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.1, | |
| "step": 1157 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1158 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "grad_norm": 0.0016401231987401843, | |
| "learning_rate": 1.2592592592592592e-05, | |
| "loss": 0.0001, | |
| "step": 1160 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1160 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1161 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1162 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1163 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1164 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1165 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1166 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1167 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1168 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "grad_norm": 0.003334516892209649, | |
| "learning_rate": 1.2222222222222222e-05, | |
| "loss": 0.0002, | |
| "step": 1170 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1170 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1171 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.11, | |
| "step": 1172 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1173 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1174 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1175 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1176 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1177 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1178 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "grad_norm": 0.004980438854545355, | |
| "learning_rate": 1.1851851851851853e-05, | |
| "loss": 0.0001, | |
| "step": 1180 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1180 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1181 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1182 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1183 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1184 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1185 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1186 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.12, | |
| "step": 1187 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1188 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "grad_norm": 0.002087895292788744, | |
| "learning_rate": 1.1481481481481482e-05, | |
| "loss": 0.0001, | |
| "step": 1190 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1190 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1191 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1192 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1193 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1194 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1195 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1196 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1197 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1198 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 4.13, | |
| "grad_norm": 0.001249711960554123, | |
| "learning_rate": 1.1111111111111112e-05, | |
| "loss": 0.0002, | |
| "step": 1200 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1200 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1201 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.13, | |
| "step": 1202 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1203 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1204 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1205 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1206 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1207 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1208 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "grad_norm": 0.0015552870463579893, | |
| "learning_rate": 1.074074074074074e-05, | |
| "loss": 0.0002, | |
| "step": 1210 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1210 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1211 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1212 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1213 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1214 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1215 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1216 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.14, | |
| "step": 1217 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1218 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "grad_norm": 0.006314919330179691, | |
| "learning_rate": 1.037037037037037e-05, | |
| "loss": 0.0001, | |
| "step": 1220 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1220 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1221 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1222 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1223 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1224 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1225 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1226 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1227 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1228 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 4.15, | |
| "grad_norm": 0.005126224830746651, | |
| "learning_rate": 1e-05, | |
| "loss": 0.0001, | |
| "step": 1230 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1230 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1231 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.15, | |
| "step": 1232 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1233 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1234 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1235 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1236 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1237 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1238 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "grad_norm": 0.0023905010893940926, | |
| "learning_rate": 9.62962962962963e-06, | |
| "loss": 0.0001, | |
| "step": 1240 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1240 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1241 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1242 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1243 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1244 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1245 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1246 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.16, | |
| "step": 1247 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1248 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "grad_norm": 0.005289968568831682, | |
| "learning_rate": 9.259259259259259e-06, | |
| "loss": 0.0007, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "eval_accuracy": 0.98, | |
| "eval_loss": 0.09618649631738663, | |
| "eval_runtime": 108.1545, | |
| "eval_samples_per_second": 0.925, | |
| "eval_steps_per_second": 0.925, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 4.17, | |
| "step": 1250 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1251 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1252 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1253 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1254 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1255 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1256 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.0, | |
| "step": 1257 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1258 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "grad_norm": 0.0012224328238517046, | |
| "learning_rate": 8.88888888888889e-06, | |
| "loss": 0.0001, | |
| "step": 1260 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1260 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1261 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1262 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1263 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1264 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1265 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1266 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1267 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1268 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "grad_norm": 0.0017797598848119378, | |
| "learning_rate": 8.518518518518519e-06, | |
| "loss": 0.0001, | |
| "step": 1270 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1270 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1271 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.01, | |
| "step": 1272 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1273 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1274 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1275 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1276 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1277 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1278 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "grad_norm": 0.024643300101161003, | |
| "learning_rate": 8.14814814814815e-06, | |
| "loss": 0.0001, | |
| "step": 1280 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1280 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1281 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1282 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1283 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1284 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1285 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1286 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.02, | |
| "step": 1287 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1288 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "grad_norm": 0.002816492458805442, | |
| "learning_rate": 7.777777777777777e-06, | |
| "loss": 0.0001, | |
| "step": 1290 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1290 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1291 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1292 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1293 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1294 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1295 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1296 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1297 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1298 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "grad_norm": 0.0015854742377996445, | |
| "learning_rate": 7.4074074074074075e-06, | |
| "loss": 0.0001, | |
| "step": 1300 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1300 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1301 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.03, | |
| "step": 1302 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1303 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1304 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1305 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1306 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1307 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1308 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "grad_norm": 0.002643193816766143, | |
| "learning_rate": 7.0370370370370375e-06, | |
| "loss": 0.0165, | |
| "step": 1310 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1310 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1311 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1312 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1313 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1314 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1315 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1316 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.04, | |
| "step": 1317 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1318 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "grad_norm": 0.002280215732753277, | |
| "learning_rate": 6.666666666666667e-06, | |
| "loss": 0.001, | |
| "step": 1320 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1320 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1321 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1322 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1323 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1324 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1325 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1326 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1327 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1328 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "grad_norm": 0.0010046822717413306, | |
| "learning_rate": 6.296296296296296e-06, | |
| "loss": 0.0001, | |
| "step": 1330 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1330 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1331 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.05, | |
| "step": 1332 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1333 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1334 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1335 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1336 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1337 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1338 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "grad_norm": 0.10451558232307434, | |
| "learning_rate": 5.925925925925927e-06, | |
| "loss": 0.0003, | |
| "step": 1340 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1340 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1341 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1342 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1343 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1344 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1345 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1346 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.06, | |
| "step": 1347 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1348 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "grad_norm": 0.006334853824228048, | |
| "learning_rate": 5.555555555555556e-06, | |
| "loss": 0.008, | |
| "step": 1350 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1350 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1351 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1352 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1353 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1354 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1355 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1356 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1357 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1358 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "grad_norm": 0.037407007068395615, | |
| "learning_rate": 5.185185185185185e-06, | |
| "loss": 0.0002, | |
| "step": 1360 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1360 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1361 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.07, | |
| "step": 1362 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1363 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1364 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1365 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1366 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1367 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1368 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "grad_norm": 0.0010384167544543743, | |
| "learning_rate": 4.814814814814815e-06, | |
| "loss": 0.0078, | |
| "step": 1370 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1370 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1371 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1372 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1373 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1374 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1375 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1376 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.08, | |
| "step": 1377 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1378 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "grad_norm": 0.0015371436020359397, | |
| "learning_rate": 4.444444444444445e-06, | |
| "loss": 0.0001, | |
| "step": 1380 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1380 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1381 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1382 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1383 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1384 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1385 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1386 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1387 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1388 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "grad_norm": 0.005509741138666868, | |
| "learning_rate": 4.074074074074075e-06, | |
| "loss": 0.0002, | |
| "step": 1390 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1390 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1391 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.09, | |
| "step": 1392 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1393 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1394 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1395 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1396 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1397 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1398 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "grad_norm": 0.0022451411932706833, | |
| "learning_rate": 3.7037037037037037e-06, | |
| "loss": 0.0002, | |
| "step": 1400 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1400 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1401 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1402 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1403 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1404 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1405 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1406 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.1, | |
| "step": 1407 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1408 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "grad_norm": 0.007799276150763035, | |
| "learning_rate": 3.3333333333333333e-06, | |
| "loss": 0.0001, | |
| "step": 1410 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1410 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1411 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1412 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1413 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1414 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1415 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1416 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1417 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1418 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "grad_norm": 0.0014103392604738474, | |
| "learning_rate": 2.9629629629629633e-06, | |
| "loss": 0.0001, | |
| "step": 1420 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1420 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1421 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.11, | |
| "step": 1422 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1423 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1424 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1425 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1426 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1427 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1428 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "grad_norm": 0.003956436179578304, | |
| "learning_rate": 2.5925925925925925e-06, | |
| "loss": 0.0001, | |
| "step": 1430 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1430 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1431 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1432 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1433 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1434 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1435 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1436 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.12, | |
| "step": 1437 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1438 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "grad_norm": 0.0046532051637768745, | |
| "learning_rate": 2.2222222222222225e-06, | |
| "loss": 0.0002, | |
| "step": 1440 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1440 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1441 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1442 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1443 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1444 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1445 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1446 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1447 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1448 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "grad_norm": 0.0016388532239943743, | |
| "learning_rate": 1.8518518518518519e-06, | |
| "loss": 0.0003, | |
| "step": 1450 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1450 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1451 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.13, | |
| "step": 1452 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1453 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1454 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1455 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1456 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1457 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1458 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "grad_norm": 1.8762030601501465, | |
| "learning_rate": 1.4814814814814817e-06, | |
| "loss": 0.0013, | |
| "step": 1460 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1460 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1461 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1462 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1463 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1464 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1465 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1466 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.14, | |
| "step": 1467 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1468 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "grad_norm": 0.0017372217262163758, | |
| "learning_rate": 1.1111111111111112e-06, | |
| "loss": 0.0002, | |
| "step": 1470 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1470 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1471 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1472 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1473 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1474 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1475 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1476 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1477 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1478 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "grad_norm": 0.001228381646797061, | |
| "learning_rate": 7.407407407407408e-07, | |
| "loss": 0.0002, | |
| "step": 1480 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1480 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1481 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.15, | |
| "step": 1482 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1483 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1484 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1485 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1486 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1487 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1488 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "grad_norm": 0.0035096406936645508, | |
| "learning_rate": 3.703703703703704e-07, | |
| "loss": 0.0001, | |
| "step": 1490 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1490 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1491 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1492 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1493 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1494 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1495 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1496 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.16, | |
| "step": 1497 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1498 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "grad_norm": 0.0030719765927642584, | |
| "learning_rate": 0.0, | |
| "loss": 0.0002, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "eval_accuracy": 0.98, | |
| "eval_loss": 0.11492405086755753, | |
| "eval_runtime": 108.1246, | |
| "eval_samples_per_second": 0.925, | |
| "eval_steps_per_second": 0.925, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "step": 1500, | |
| "total_flos": 1.64282826710016e+18, | |
| "train_loss": 0.16299300388417517, | |
| "train_runtime": 2410.5738, | |
| "train_samples_per_second": 0.622, | |
| "train_steps_per_second": 0.622 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "eval_accuracy": 0.97, | |
| "eval_loss": 0.10209280252456665, | |
| "eval_runtime": 111.5723, | |
| "eval_samples_per_second": 0.896, | |
| "eval_steps_per_second": 0.896, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 0.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "accuracy": 1.0, | |
| "epoch": 5.17, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "eval_accuracy": 0.97, | |
| "eval_loss": 0.10209278017282486, | |
| "eval_runtime": 111.8728, | |
| "eval_samples_per_second": 0.894, | |
| "eval_steps_per_second": 0.894, | |
| "step": 1500 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 1500, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9223372036854775807, | |
| "save_steps": 500, | |
| "total_flos": 1.64282826710016e+18, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |