Commit ·
92f644a
1
Parent(s): eb982a5
Upload trainer_state.json with huggingface_hub
Browse files- trainer_state.json +142 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,142 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.2659623324871063,
|
| 3 |
+
"best_model_checkpoint": "./models/results_semeval_2017_task_7_453/checkpoint-40",
|
| 4 |
+
"epoch": 0.9078014184397163,
|
| 5 |
+
"global_step": 40,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.11,
|
| 12 |
+
"eval_accuracy": 0.7047146401985112,
|
| 13 |
+
"eval_f1": 0.826783114992722,
|
| 14 |
+
"eval_loss": 0.5836203098297119,
|
| 15 |
+
"eval_precision": 0.7047146401985112,
|
| 16 |
+
"eval_recall": 1.0,
|
| 17 |
+
"eval_runtime": 2.1688,
|
| 18 |
+
"eval_samples_per_second": 185.814,
|
| 19 |
+
"eval_steps_per_second": 23.515,
|
| 20 |
+
"step": 5
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.18,
|
| 24 |
+
"learning_rate": 4.0909090909090915e-05,
|
| 25 |
+
"loss": 0.6065,
|
| 26 |
+
"step": 8
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.23,
|
| 30 |
+
"eval_accuracy": 0.7047146401985112,
|
| 31 |
+
"eval_f1": 0.826783114992722,
|
| 32 |
+
"eval_loss": 0.5035943984985352,
|
| 33 |
+
"eval_precision": 0.7047146401985112,
|
| 34 |
+
"eval_recall": 1.0,
|
| 35 |
+
"eval_runtime": 2.1003,
|
| 36 |
+
"eval_samples_per_second": 191.876,
|
| 37 |
+
"eval_steps_per_second": 24.282,
|
| 38 |
+
"step": 10
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.34,
|
| 42 |
+
"eval_accuracy": 0.7047146401985112,
|
| 43 |
+
"eval_f1": 0.826783114992722,
|
| 44 |
+
"eval_loss": 0.4322372078895569,
|
| 45 |
+
"eval_precision": 0.7047146401985112,
|
| 46 |
+
"eval_recall": 1.0,
|
| 47 |
+
"eval_runtime": 2.1086,
|
| 48 |
+
"eval_samples_per_second": 191.125,
|
| 49 |
+
"eval_steps_per_second": 24.187,
|
| 50 |
+
"step": 15
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.36,
|
| 54 |
+
"learning_rate": 3.181818181818182e-05,
|
| 55 |
+
"loss": 0.5295,
|
| 56 |
+
"step": 16
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.45,
|
| 60 |
+
"eval_accuracy": 0.7047146401985112,
|
| 61 |
+
"eval_f1": 0.826783114992722,
|
| 62 |
+
"eval_loss": 0.4136073887348175,
|
| 63 |
+
"eval_precision": 0.7047146401985112,
|
| 64 |
+
"eval_recall": 1.0,
|
| 65 |
+
"eval_runtime": 2.1323,
|
| 66 |
+
"eval_samples_per_second": 188.994,
|
| 67 |
+
"eval_steps_per_second": 23.917,
|
| 68 |
+
"step": 20
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.54,
|
| 72 |
+
"learning_rate": 2.272727272727273e-05,
|
| 73 |
+
"loss": 0.4569,
|
| 74 |
+
"step": 24
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 0.57,
|
| 78 |
+
"eval_accuracy": 0.8759305210918115,
|
| 79 |
+
"eval_f1": 0.9113475177304964,
|
| 80 |
+
"eval_loss": 0.3490051329135895,
|
| 81 |
+
"eval_precision": 0.9178571428571428,
|
| 82 |
+
"eval_recall": 0.9049295774647887,
|
| 83 |
+
"eval_runtime": 2.1172,
|
| 84 |
+
"eval_samples_per_second": 190.346,
|
| 85 |
+
"eval_steps_per_second": 24.088,
|
| 86 |
+
"step": 25
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.68,
|
| 90 |
+
"eval_accuracy": 0.8660049627791563,
|
| 91 |
+
"eval_f1": 0.9096989966555185,
|
| 92 |
+
"eval_loss": 0.3699808120727539,
|
| 93 |
+
"eval_precision": 0.8662420382165605,
|
| 94 |
+
"eval_recall": 0.9577464788732394,
|
| 95 |
+
"eval_runtime": 2.1186,
|
| 96 |
+
"eval_samples_per_second": 190.215,
|
| 97 |
+
"eval_steps_per_second": 24.072,
|
| 98 |
+
"step": 30
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.73,
|
| 102 |
+
"learning_rate": 1.3636363636363637e-05,
|
| 103 |
+
"loss": 0.3281,
|
| 104 |
+
"step": 32
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 0.79,
|
| 108 |
+
"eval_accuracy": 0.8833746898263027,
|
| 109 |
+
"eval_f1": 0.9182608695652175,
|
| 110 |
+
"eval_loss": 0.3029639422893524,
|
| 111 |
+
"eval_precision": 0.9072164948453608,
|
| 112 |
+
"eval_recall": 0.9295774647887324,
|
| 113 |
+
"eval_runtime": 2.1093,
|
| 114 |
+
"eval_samples_per_second": 191.061,
|
| 115 |
+
"eval_steps_per_second": 24.179,
|
| 116 |
+
"step": 35
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 0.91,
|
| 120 |
+
"learning_rate": 4.5454545454545455e-06,
|
| 121 |
+
"loss": 0.288,
|
| 122 |
+
"step": 40
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.91,
|
| 126 |
+
"eval_accuracy": 0.8858560794044665,
|
| 127 |
+
"eval_f1": 0.9172661870503597,
|
| 128 |
+
"eval_loss": 0.2659623324871063,
|
| 129 |
+
"eval_precision": 0.9375,
|
| 130 |
+
"eval_recall": 0.897887323943662,
|
| 131 |
+
"eval_runtime": 2.1394,
|
| 132 |
+
"eval_samples_per_second": 188.37,
|
| 133 |
+
"eval_steps_per_second": 23.838,
|
| 134 |
+
"step": 40
|
| 135 |
+
}
|
| 136 |
+
],
|
| 137 |
+
"max_steps": 44,
|
| 138 |
+
"num_train_epochs": 1,
|
| 139 |
+
"total_flos": 111822198528000.0,
|
| 140 |
+
"trial_name": null,
|
| 141 |
+
"trial_params": null
|
| 142 |
+
}
|