Delete trainer_state.json
Browse files- trainer_state.json +0 -152
trainer_state.json
DELETED
|
@@ -1,152 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": 0.19506023824214935,
|
| 3 |
-
"best_model_checkpoint": "bioformer16L_caption_RTX6000/checkpoint-1230",
|
| 4 |
-
"epoch": 4.0,
|
| 5 |
-
"eval_steps": 500,
|
| 6 |
-
"global_step": 4920,
|
| 7 |
-
"is_hyper_param_search": false,
|
| 8 |
-
"is_local_process_zero": true,
|
| 9 |
-
"is_world_process_zero": true,
|
| 10 |
-
"log_history": [
|
| 11 |
-
{
|
| 12 |
-
"epoch": 0.4065040650406504,
|
| 13 |
-
"grad_norm": 4.495114803314209,
|
| 14 |
-
"learning_rate": 2.75609756097561e-05,
|
| 15 |
-
"loss": 0.256,
|
| 16 |
-
"step": 500
|
| 17 |
-
},
|
| 18 |
-
{
|
| 19 |
-
"epoch": 0.8130081300813008,
|
| 20 |
-
"grad_norm": 2.5886974334716797,
|
| 21 |
-
"learning_rate": 2.5121951219512197e-05,
|
| 22 |
-
"loss": 0.1859,
|
| 23 |
-
"step": 1000
|
| 24 |
-
},
|
| 25 |
-
{
|
| 26 |
-
"epoch": 1.0,
|
| 27 |
-
"eval_accuracy": 0.9271035681423168,
|
| 28 |
-
"eval_f1": 0.9220692013036642,
|
| 29 |
-
"eval_f1_medical": 0.9418765794407761,
|
| 30 |
-
"eval_f1_non_medical": 0.9022618231665525,
|
| 31 |
-
"eval_loss": 0.19506023824214935,
|
| 32 |
-
"eval_precision": 0.9249732755411849,
|
| 33 |
-
"eval_recall": 0.9194829650061407,
|
| 34 |
-
"eval_runtime": 34.6825,
|
| 35 |
-
"eval_samples_per_second": 282.015,
|
| 36 |
-
"eval_steps_per_second": 8.823,
|
| 37 |
-
"step": 1230
|
| 38 |
-
},
|
| 39 |
-
{
|
| 40 |
-
"epoch": 1.2195121951219512,
|
| 41 |
-
"grad_norm": 2.54653263092041,
|
| 42 |
-
"learning_rate": 2.2682926829268295e-05,
|
| 43 |
-
"loss": 0.1571,
|
| 44 |
-
"step": 1500
|
| 45 |
-
},
|
| 46 |
-
{
|
| 47 |
-
"epoch": 1.6260162601626016,
|
| 48 |
-
"grad_norm": 2.3185017108917236,
|
| 49 |
-
"learning_rate": 2.024390243902439e-05,
|
| 50 |
-
"loss": 0.1303,
|
| 51 |
-
"step": 2000
|
| 52 |
-
},
|
| 53 |
-
{
|
| 54 |
-
"epoch": 2.0,
|
| 55 |
-
"eval_accuracy": 0.92516102647991,
|
| 56 |
-
"eval_f1": 0.921045875731096,
|
| 57 |
-
"eval_f1_medical": 0.9390710837356417,
|
| 58 |
-
"eval_f1_non_medical": 0.9030206677265501,
|
| 59 |
-
"eval_loss": 0.2094709575176239,
|
| 60 |
-
"eval_precision": 0.9184306018002388,
|
| 61 |
-
"eval_recall": 0.9240747625361165,
|
| 62 |
-
"eval_runtime": 34.5371,
|
| 63 |
-
"eval_samples_per_second": 283.202,
|
| 64 |
-
"eval_steps_per_second": 8.86,
|
| 65 |
-
"step": 2460
|
| 66 |
-
},
|
| 67 |
-
{
|
| 68 |
-
"epoch": 2.032520325203252,
|
| 69 |
-
"grad_norm": 4.058623313903809,
|
| 70 |
-
"learning_rate": 1.7804878048780488e-05,
|
| 71 |
-
"loss": 0.1238,
|
| 72 |
-
"step": 2500
|
| 73 |
-
},
|
| 74 |
-
{
|
| 75 |
-
"epoch": 2.4390243902439024,
|
| 76 |
-
"grad_norm": 5.887447834014893,
|
| 77 |
-
"learning_rate": 1.5365853658536586e-05,
|
| 78 |
-
"loss": 0.0884,
|
| 79 |
-
"step": 3000
|
| 80 |
-
},
|
| 81 |
-
{
|
| 82 |
-
"epoch": 2.845528455284553,
|
| 83 |
-
"grad_norm": 3.797337055206299,
|
| 84 |
-
"learning_rate": 1.2926829268292684e-05,
|
| 85 |
-
"loss": 0.092,
|
| 86 |
-
"step": 3500
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 3.0,
|
| 90 |
-
"eval_accuracy": 0.9272058071771803,
|
| 91 |
-
"eval_f1": 0.9219373155502308,
|
| 92 |
-
"eval_f1_medical": 0.9422171725369258,
|
| 93 |
-
"eval_f1_non_medical": 0.901657458563536,
|
| 94 |
-
"eval_loss": 0.27733102440834045,
|
| 95 |
-
"eval_precision": 0.9264226794905697,
|
| 96 |
-
"eval_recall": 0.9181447502548421,
|
| 97 |
-
"eval_runtime": 34.4489,
|
| 98 |
-
"eval_samples_per_second": 283.927,
|
| 99 |
-
"eval_steps_per_second": 8.883,
|
| 100 |
-
"step": 3690
|
| 101 |
-
},
|
| 102 |
-
{
|
| 103 |
-
"epoch": 3.252032520325203,
|
| 104 |
-
"grad_norm": 11.003833770751953,
|
| 105 |
-
"learning_rate": 1.048780487804878e-05,
|
| 106 |
-
"loss": 0.0625,
|
| 107 |
-
"step": 4000
|
| 108 |
-
},
|
| 109 |
-
{
|
| 110 |
-
"epoch": 3.658536585365854,
|
| 111 |
-
"grad_norm": 11.000012397766113,
|
| 112 |
-
"learning_rate": 8.048780487804879e-06,
|
| 113 |
-
"loss": 0.059,
|
| 114 |
-
"step": 4500
|
| 115 |
-
},
|
| 116 |
-
{
|
| 117 |
-
"epoch": 4.0,
|
| 118 |
-
"eval_accuracy": 0.9267968510377262,
|
| 119 |
-
"eval_f1": 0.9224128655389552,
|
| 120 |
-
"eval_f1_medical": 0.9408557739963654,
|
| 121 |
-
"eval_f1_non_medical": 0.903969957081545,
|
| 122 |
-
"eval_loss": 0.30652934312820435,
|
| 123 |
-
"eval_precision": 0.9214772563975018,
|
| 124 |
-
"eval_recall": 0.9233923856098107,
|
| 125 |
-
"eval_runtime": 34.3959,
|
| 126 |
-
"eval_samples_per_second": 284.366,
|
| 127 |
-
"eval_steps_per_second": 8.896,
|
| 128 |
-
"step": 4920
|
| 129 |
-
}
|
| 130 |
-
],
|
| 131 |
-
"logging_steps": 500,
|
| 132 |
-
"max_steps": 6150,
|
| 133 |
-
"num_input_tokens_seen": 0,
|
| 134 |
-
"num_train_epochs": 5,
|
| 135 |
-
"save_steps": 500,
|
| 136 |
-
"stateful_callbacks": {
|
| 137 |
-
"TrainerControl": {
|
| 138 |
-
"args": {
|
| 139 |
-
"should_epoch_stop": false,
|
| 140 |
-
"should_evaluate": false,
|
| 141 |
-
"should_log": false,
|
| 142 |
-
"should_save": true,
|
| 143 |
-
"should_training_stop": false
|
| 144 |
-
},
|
| 145 |
-
"attributes": {}
|
| 146 |
-
}
|
| 147 |
-
},
|
| 148 |
-
"total_flos": 1.1766206373267072e+16,
|
| 149 |
-
"train_batch_size": 32,
|
| 150 |
-
"trial_name": null,
|
| 151 |
-
"trial_params": null
|
| 152 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|