Upload trainer_state.json
Browse files- trainer_state.json +88 -0
trainer_state.json
ADDED
|
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 29.41176470588235,
|
| 5 |
+
"global_step": 1000,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 5.88,
|
| 12 |
+
"eval_accuracy": 0.6521739130434783,
|
| 13 |
+
"eval_f1": 0.5579710144927535,
|
| 14 |
+
"eval_loss": 1.4428606033325195,
|
| 15 |
+
"eval_precision": 0.5144927536231884,
|
| 16 |
+
"eval_recall": 0.6521739130434783,
|
| 17 |
+
"eval_runtime": 0.6869,
|
| 18 |
+
"eval_samples_per_second": 33.482,
|
| 19 |
+
"eval_steps_per_second": 17.469,
|
| 20 |
+
"step": 200
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 11.76,
|
| 24 |
+
"eval_accuracy": 1.0,
|
| 25 |
+
"eval_f1": 1.0,
|
| 26 |
+
"eval_loss": 0.2573736310005188,
|
| 27 |
+
"eval_precision": 1.0,
|
| 28 |
+
"eval_recall": 1.0,
|
| 29 |
+
"eval_runtime": 0.6853,
|
| 30 |
+
"eval_samples_per_second": 33.564,
|
| 31 |
+
"eval_steps_per_second": 17.512,
|
| 32 |
+
"step": 400
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 14.71,
|
| 36 |
+
"learning_rate": 3.529411764705883e-05,
|
| 37 |
+
"loss": 1.1618,
|
| 38 |
+
"step": 500
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 17.65,
|
| 42 |
+
"eval_accuracy": 1.0,
|
| 43 |
+
"eval_f1": 1.0,
|
| 44 |
+
"eval_loss": 0.048190146684646606,
|
| 45 |
+
"eval_precision": 1.0,
|
| 46 |
+
"eval_recall": 1.0,
|
| 47 |
+
"eval_runtime": 0.6929,
|
| 48 |
+
"eval_samples_per_second": 33.194,
|
| 49 |
+
"eval_steps_per_second": 17.319,
|
| 50 |
+
"step": 600
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 23.53,
|
| 54 |
+
"eval_accuracy": 1.0,
|
| 55 |
+
"eval_f1": 1.0,
|
| 56 |
+
"eval_loss": 0.020614268258213997,
|
| 57 |
+
"eval_precision": 1.0,
|
| 58 |
+
"eval_recall": 1.0,
|
| 59 |
+
"eval_runtime": 0.702,
|
| 60 |
+
"eval_samples_per_second": 32.762,
|
| 61 |
+
"eval_steps_per_second": 17.093,
|
| 62 |
+
"step": 800
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 29.41,
|
| 66 |
+
"learning_rate": 2.058823529411765e-05,
|
| 67 |
+
"loss": 0.0105,
|
| 68 |
+
"step": 1000
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 29.41,
|
| 72 |
+
"eval_accuracy": 1.0,
|
| 73 |
+
"eval_f1": 1.0,
|
| 74 |
+
"eval_loss": 0.014327215030789375,
|
| 75 |
+
"eval_precision": 1.0,
|
| 76 |
+
"eval_recall": 1.0,
|
| 77 |
+
"eval_runtime": 0.6944,
|
| 78 |
+
"eval_samples_per_second": 33.122,
|
| 79 |
+
"eval_steps_per_second": 17.281,
|
| 80 |
+
"step": 1000
|
| 81 |
+
}
|
| 82 |
+
],
|
| 83 |
+
"max_steps": 1700,
|
| 84 |
+
"num_train_epochs": 50,
|
| 85 |
+
"total_flos": 518689670888448.0,
|
| 86 |
+
"trial_name": null,
|
| 87 |
+
"trial_params": null
|
| 88 |
+
}
|