Commit ·
f5037fc
1
Parent(s): 3b09303
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": 0.2773796319961548,
|
| 3 |
+
"best_model_checkpoint": "./models/results_semeval_2017_task_7_693/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": 0.5785936117172241,
|
| 15 |
+
"eval_precision": 0.7047146401985112,
|
| 16 |
+
"eval_recall": 1.0,
|
| 17 |
+
"eval_runtime": 2.1163,
|
| 18 |
+
"eval_samples_per_second": 190.428,
|
| 19 |
+
"eval_steps_per_second": 24.099,
|
| 20 |
+
"step": 5
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.18,
|
| 24 |
+
"learning_rate": 4.0909090909090915e-05,
|
| 25 |
+
"loss": 0.6121,
|
| 26 |
+
"step": 8
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.23,
|
| 30 |
+
"eval_accuracy": 0.7047146401985112,
|
| 31 |
+
"eval_f1": 0.826783114992722,
|
| 32 |
+
"eval_loss": 0.512681782245636,
|
| 33 |
+
"eval_precision": 0.7047146401985112,
|
| 34 |
+
"eval_recall": 1.0,
|
| 35 |
+
"eval_runtime": 2.1906,
|
| 36 |
+
"eval_samples_per_second": 183.972,
|
| 37 |
+
"eval_steps_per_second": 23.282,
|
| 38 |
+
"step": 10
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.34,
|
| 42 |
+
"eval_accuracy": 0.7047146401985112,
|
| 43 |
+
"eval_f1": 0.826783114992722,
|
| 44 |
+
"eval_loss": 0.41043248772621155,
|
| 45 |
+
"eval_precision": 0.7047146401985112,
|
| 46 |
+
"eval_recall": 1.0,
|
| 47 |
+
"eval_runtime": 2.1431,
|
| 48 |
+
"eval_samples_per_second": 188.049,
|
| 49 |
+
"eval_steps_per_second": 23.798,
|
| 50 |
+
"step": 15
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.36,
|
| 54 |
+
"learning_rate": 3.181818181818182e-05,
|
| 55 |
+
"loss": 0.5116,
|
| 56 |
+
"step": 16
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.45,
|
| 60 |
+
"eval_accuracy": 0.7568238213399504,
|
| 61 |
+
"eval_f1": 0.8515151515151514,
|
| 62 |
+
"eval_loss": 0.37497952580451965,
|
| 63 |
+
"eval_precision": 0.7473404255319149,
|
| 64 |
+
"eval_recall": 0.9894366197183099,
|
| 65 |
+
"eval_runtime": 2.1596,
|
| 66 |
+
"eval_samples_per_second": 186.61,
|
| 67 |
+
"eval_steps_per_second": 23.616,
|
| 68 |
+
"step": 20
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.54,
|
| 72 |
+
"learning_rate": 2.272727272727273e-05,
|
| 73 |
+
"loss": 0.3673,
|
| 74 |
+
"step": 24
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 0.57,
|
| 78 |
+
"eval_accuracy": 0.8883374689826302,
|
| 79 |
+
"eval_f1": 0.9189189189189191,
|
| 80 |
+
"eval_loss": 0.28403061628341675,
|
| 81 |
+
"eval_precision": 0.940959409594096,
|
| 82 |
+
"eval_recall": 0.897887323943662,
|
| 83 |
+
"eval_runtime": 2.1799,
|
| 84 |
+
"eval_samples_per_second": 184.873,
|
| 85 |
+
"eval_steps_per_second": 23.396,
|
| 86 |
+
"step": 25
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.68,
|
| 90 |
+
"eval_accuracy": 0.8957816377171216,
|
| 91 |
+
"eval_f1": 0.9275862068965517,
|
| 92 |
+
"eval_loss": 0.3391099274158478,
|
| 93 |
+
"eval_precision": 0.9087837837837838,
|
| 94 |
+
"eval_recall": 0.9471830985915493,
|
| 95 |
+
"eval_runtime": 2.1847,
|
| 96 |
+
"eval_samples_per_second": 184.465,
|
| 97 |
+
"eval_steps_per_second": 23.344,
|
| 98 |
+
"step": 30
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.73,
|
| 102 |
+
"learning_rate": 1.4772727272727274e-05,
|
| 103 |
+
"loss": 0.3116,
|
| 104 |
+
"step": 32
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 0.79,
|
| 108 |
+
"eval_accuracy": 0.890818858560794,
|
| 109 |
+
"eval_f1": 0.9230769230769231,
|
| 110 |
+
"eval_loss": 0.31251412630081177,
|
| 111 |
+
"eval_precision": 0.9166666666666666,
|
| 112 |
+
"eval_recall": 0.9295774647887324,
|
| 113 |
+
"eval_runtime": 2.1117,
|
| 114 |
+
"eval_samples_per_second": 190.845,
|
| 115 |
+
"eval_steps_per_second": 24.152,
|
| 116 |
+
"step": 35
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 0.91,
|
| 120 |
+
"learning_rate": 5.681818181818182e-06,
|
| 121 |
+
"loss": 0.2684,
|
| 122 |
+
"step": 40
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.91,
|
| 126 |
+
"eval_accuracy": 0.890818858560794,
|
| 127 |
+
"eval_f1": 0.9225352112676056,
|
| 128 |
+
"eval_loss": 0.2773796319961548,
|
| 129 |
+
"eval_precision": 0.9225352112676056,
|
| 130 |
+
"eval_recall": 0.9225352112676056,
|
| 131 |
+
"eval_runtime": 2.1762,
|
| 132 |
+
"eval_samples_per_second": 185.184,
|
| 133 |
+
"eval_steps_per_second": 23.435,
|
| 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 |
+
}
|