End of training
Browse files- all_results.json +17 -0
- eval_results.json +12 -0
- train_results.json +8 -0
- trainer_state.json +1105 -0
all_results.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_accuracy": 0.9926599431486313,
|
| 4 |
+
"eval_f1": 0.9590143324113654,
|
| 5 |
+
"eval_loss": 0.0412251353263855,
|
| 6 |
+
"eval_precision": 0.9552512940390716,
|
| 7 |
+
"eval_recall": 0.9628071356445641,
|
| 8 |
+
"eval_runtime": 6.2543,
|
| 9 |
+
"eval_samples": 3251,
|
| 10 |
+
"eval_samples_per_second": 519.806,
|
| 11 |
+
"eval_steps_per_second": 0.64,
|
| 12 |
+
"train_loss": 0.022697925716638565,
|
| 13 |
+
"train_runtime": 1311.4885,
|
| 14 |
+
"train_samples": 14042,
|
| 15 |
+
"train_samples_per_second": 107.069,
|
| 16 |
+
"train_steps_per_second": 1.677
|
| 17 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_accuracy": 0.9926599431486313,
|
| 4 |
+
"eval_f1": 0.9590143324113654,
|
| 5 |
+
"eval_loss": 0.0412251353263855,
|
| 6 |
+
"eval_precision": 0.9552512940390716,
|
| 7 |
+
"eval_recall": 0.9628071356445641,
|
| 8 |
+
"eval_runtime": 6.2543,
|
| 9 |
+
"eval_samples": 3251,
|
| 10 |
+
"eval_samples_per_second": 519.806,
|
| 11 |
+
"eval_steps_per_second": 0.64
|
| 12 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"train_loss": 0.022697925716638565,
|
| 4 |
+
"train_runtime": 1311.4885,
|
| 5 |
+
"train_samples": 14042,
|
| 6 |
+
"train_samples_per_second": 107.069,
|
| 7 |
+
"train_steps_per_second": 1.677
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.9590143324113654,
|
| 3 |
+
"best_model_checkpoint": "twitter-roberta-base-dec2021-CoNLL/checkpoint-2000",
|
| 4 |
+
"epoch": 10.0,
|
| 5 |
+
"global_step": 2200,
|
| 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": 0.9349324403255325,
|
| 13 |
+
"eval_f1": 0.5845156059991893,
|
| 14 |
+
"eval_loss": 0.21257123351097107,
|
| 15 |
+
"eval_precision": 0.5638980134522134,
|
| 16 |
+
"eval_recall": 0.6066980814540559,
|
| 17 |
+
"eval_runtime": 5.9566,
|
| 18 |
+
"eval_samples_per_second": 545.785,
|
| 19 |
+
"eval_steps_per_second": 0.672,
|
| 20 |
+
"step": 25
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.23,
|
| 24 |
+
"eval_accuracy": 0.9765001362875277,
|
| 25 |
+
"eval_f1": 0.8431372549019608,
|
| 26 |
+
"eval_loss": 0.08492632955312729,
|
| 27 |
+
"eval_precision": 0.8258553905745643,
|
| 28 |
+
"eval_recall": 0.8611578593066308,
|
| 29 |
+
"eval_runtime": 5.965,
|
| 30 |
+
"eval_samples_per_second": 545.015,
|
| 31 |
+
"eval_steps_per_second": 0.671,
|
| 32 |
+
"step": 50
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.34,
|
| 36 |
+
"eval_accuracy": 0.9819905766909388,
|
| 37 |
+
"eval_f1": 0.8853031689262247,
|
| 38 |
+
"eval_loss": 0.06398079544305801,
|
| 39 |
+
"eval_precision": 0.8751850024666995,
|
| 40 |
+
"eval_recall": 0.8956580276001347,
|
| 41 |
+
"eval_runtime": 6.2595,
|
| 42 |
+
"eval_samples_per_second": 519.369,
|
| 43 |
+
"eval_steps_per_second": 0.639,
|
| 44 |
+
"step": 75
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.45,
|
| 48 |
+
"eval_accuracy": 0.9831587555001752,
|
| 49 |
+
"eval_f1": 0.8939961695395121,
|
| 50 |
+
"eval_loss": 0.05720128118991852,
|
| 51 |
+
"eval_precision": 0.88478655018955,
|
| 52 |
+
"eval_recall": 0.9033995287781892,
|
| 53 |
+
"eval_runtime": 6.1814,
|
| 54 |
+
"eval_samples_per_second": 525.936,
|
| 55 |
+
"eval_steps_per_second": 0.647,
|
| 56 |
+
"step": 100
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.57,
|
| 60 |
+
"eval_accuracy": 0.986585413340602,
|
| 61 |
+
"eval_f1": 0.9154577288644322,
|
| 62 |
+
"eval_loss": 0.046862851828336716,
|
| 63 |
+
"eval_precision": 0.9071381361533377,
|
| 64 |
+
"eval_recall": 0.9239313362504208,
|
| 65 |
+
"eval_runtime": 6.2511,
|
| 66 |
+
"eval_samples_per_second": 520.07,
|
| 67 |
+
"eval_steps_per_second": 0.64,
|
| 68 |
+
"step": 125
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.68,
|
| 72 |
+
"eval_accuracy": 0.987675713562556,
|
| 73 |
+
"eval_f1": 0.9237600536193029,
|
| 74 |
+
"eval_loss": 0.044157467782497406,
|
| 75 |
+
"eval_precision": 0.9197530864197531,
|
| 76 |
+
"eval_recall": 0.927802086839448,
|
| 77 |
+
"eval_runtime": 5.9975,
|
| 78 |
+
"eval_samples_per_second": 542.063,
|
| 79 |
+
"eval_steps_per_second": 0.667,
|
| 80 |
+
"step": 150
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.8,
|
| 84 |
+
"eval_accuracy": 0.9880845761457887,
|
| 85 |
+
"eval_f1": 0.9256350267379679,
|
| 86 |
+
"eval_loss": 0.04243418201804161,
|
| 87 |
+
"eval_precision": 0.9191835380019914,
|
| 88 |
+
"eval_recall": 0.9321777179400875,
|
| 89 |
+
"eval_runtime": 6.2444,
|
| 90 |
+
"eval_samples_per_second": 520.63,
|
| 91 |
+
"eval_steps_per_second": 0.641,
|
| 92 |
+
"step": 175
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.91,
|
| 96 |
+
"eval_accuracy": 0.9890775281336397,
|
| 97 |
+
"eval_f1": 0.9290815479156287,
|
| 98 |
+
"eval_loss": 0.04066213220357895,
|
| 99 |
+
"eval_precision": 0.9170491803278689,
|
| 100 |
+
"eval_recall": 0.9414338606529788,
|
| 101 |
+
"eval_runtime": 6.0302,
|
| 102 |
+
"eval_samples_per_second": 539.124,
|
| 103 |
+
"eval_steps_per_second": 0.663,
|
| 104 |
+
"step": 200
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 1.02,
|
| 108 |
+
"eval_accuracy": 0.9894279817764106,
|
| 109 |
+
"eval_f1": 0.9332665163284056,
|
| 110 |
+
"eval_loss": 0.0402386300265789,
|
| 111 |
+
"eval_precision": 0.9263803680981595,
|
| 112 |
+
"eval_recall": 0.9402558061258836,
|
| 113 |
+
"eval_runtime": 6.3589,
|
| 114 |
+
"eval_samples_per_second": 511.254,
|
| 115 |
+
"eval_steps_per_second": 0.629,
|
| 116 |
+
"step": 225
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.14,
|
| 120 |
+
"eval_accuracy": 0.9897394961255402,
|
| 121 |
+
"eval_f1": 0.9387072497700476,
|
| 122 |
+
"eval_loss": 0.03986027464270592,
|
| 123 |
+
"eval_precision": 0.932856905434602,
|
| 124 |
+
"eval_recall": 0.9446314372265231,
|
| 125 |
+
"eval_runtime": 6.1252,
|
| 126 |
+
"eval_samples_per_second": 530.756,
|
| 127 |
+
"eval_steps_per_second": 0.653,
|
| 128 |
+
"step": 250
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 1.25,
|
| 132 |
+
"eval_accuracy": 0.9896616175382579,
|
| 133 |
+
"eval_f1": 0.9345029239766081,
|
| 134 |
+
"eval_loss": 0.038380175828933716,
|
| 135 |
+
"eval_precision": 0.9278367617783676,
|
| 136 |
+
"eval_recall": 0.941265567149108,
|
| 137 |
+
"eval_runtime": 6.065,
|
| 138 |
+
"eval_samples_per_second": 536.028,
|
| 139 |
+
"eval_steps_per_second": 0.66,
|
| 140 |
+
"step": 275
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 1.36,
|
| 144 |
+
"eval_accuracy": 0.9905572212920057,
|
| 145 |
+
"eval_f1": 0.9427423405323957,
|
| 146 |
+
"eval_loss": 0.03632938861846924,
|
| 147 |
+
"eval_precision": 0.9378747501665556,
|
| 148 |
+
"eval_recall": 0.9476607202961965,
|
| 149 |
+
"eval_runtime": 6.3308,
|
| 150 |
+
"eval_samples_per_second": 513.519,
|
| 151 |
+
"eval_steps_per_second": 0.632,
|
| 152 |
+
"step": 300
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 1.48,
|
| 156 |
+
"eval_accuracy": 0.9904598730579027,
|
| 157 |
+
"eval_f1": 0.9436266309802609,
|
| 158 |
+
"eval_loss": 0.03615300729870796,
|
| 159 |
+
"eval_precision": 0.9379780512138344,
|
| 160 |
+
"eval_recall": 0.9493436553349041,
|
| 161 |
+
"eval_runtime": 6.0859,
|
| 162 |
+
"eval_samples_per_second": 534.186,
|
| 163 |
+
"eval_steps_per_second": 0.657,
|
| 164 |
+
"step": 325
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 1.59,
|
| 168 |
+
"eval_accuracy": 0.9905377516451852,
|
| 169 |
+
"eval_f1": 0.9446723026701265,
|
| 170 |
+
"eval_loss": 0.03641640394926071,
|
| 171 |
+
"eval_precision": 0.9397169025811823,
|
| 172 |
+
"eval_recall": 0.9496802423426456,
|
| 173 |
+
"eval_runtime": 6.3707,
|
| 174 |
+
"eval_samples_per_second": 510.303,
|
| 175 |
+
"eval_steps_per_second": 0.628,
|
| 176 |
+
"step": 350
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 1.7,
|
| 180 |
+
"eval_accuracy": 0.9898952533001051,
|
| 181 |
+
"eval_f1": 0.9398998330550917,
|
| 182 |
+
"eval_loss": 0.03667406737804413,
|
| 183 |
+
"eval_precision": 0.9324279562769129,
|
| 184 |
+
"eval_recall": 0.9474924267923258,
|
| 185 |
+
"eval_runtime": 6.0794,
|
| 186 |
+
"eval_samples_per_second": 534.757,
|
| 187 |
+
"eval_steps_per_second": 0.658,
|
| 188 |
+
"step": 375
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 1.82,
|
| 192 |
+
"eval_accuracy": 0.9898952533001051,
|
| 193 |
+
"eval_f1": 0.9404383469968212,
|
| 194 |
+
"eval_loss": 0.03719446063041687,
|
| 195 |
+
"eval_precision": 0.9349634065202928,
|
| 196 |
+
"eval_recall": 0.9459777852574891,
|
| 197 |
+
"eval_runtime": 6.4259,
|
| 198 |
+
"eval_samples_per_second": 505.925,
|
| 199 |
+
"eval_steps_per_second": 0.622,
|
| 200 |
+
"step": 400
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.93,
|
| 204 |
+
"eval_accuracy": 0.9909076749347767,
|
| 205 |
+
"eval_f1": 0.9461879655201272,
|
| 206 |
+
"eval_loss": 0.033916592597961426,
|
| 207 |
+
"eval_precision": 0.9410687531213584,
|
| 208 |
+
"eval_recall": 0.9513631773813531,
|
| 209 |
+
"eval_runtime": 6.1497,
|
| 210 |
+
"eval_samples_per_second": 528.643,
|
| 211 |
+
"eval_steps_per_second": 0.65,
|
| 212 |
+
"step": 425
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 2.05,
|
| 216 |
+
"eval_accuracy": 0.9911413106966239,
|
| 217 |
+
"eval_f1": 0.9473772274742742,
|
| 218 |
+
"eval_loss": 0.03361179679632187,
|
| 219 |
+
"eval_precision": 0.9419397770753618,
|
| 220 |
+
"eval_recall": 0.9528778189161898,
|
| 221 |
+
"eval_runtime": 6.1538,
|
| 222 |
+
"eval_samples_per_second": 528.288,
|
| 223 |
+
"eval_steps_per_second": 0.65,
|
| 224 |
+
"step": 450
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 2.16,
|
| 228 |
+
"eval_accuracy": 0.9913554768116506,
|
| 229 |
+
"eval_f1": 0.9491667364542332,
|
| 230 |
+
"eval_loss": 0.03356914222240448,
|
| 231 |
+
"eval_precision": 0.9446574429071511,
|
| 232 |
+
"eval_recall": 0.9537192864355436,
|
| 233 |
+
"eval_runtime": 6.3848,
|
| 234 |
+
"eval_samples_per_second": 509.18,
|
| 235 |
+
"eval_steps_per_second": 0.626,
|
| 236 |
+
"step": 475
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 2.27,
|
| 240 |
+
"learning_rate": 3.8636363636363636e-05,
|
| 241 |
+
"loss": 0.079,
|
| 242 |
+
"step": 500
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 2.27,
|
| 246 |
+
"eval_accuracy": 0.9913749464584712,
|
| 247 |
+
"eval_f1": 0.9492317969271877,
|
| 248 |
+
"eval_loss": 0.03446485847234726,
|
| 249 |
+
"eval_precision": 0.9419953596287703,
|
| 250 |
+
"eval_recall": 0.9565802760013463,
|
| 251 |
+
"eval_runtime": 6.1351,
|
| 252 |
+
"eval_samples_per_second": 529.902,
|
| 253 |
+
"eval_steps_per_second": 0.652,
|
| 254 |
+
"step": 500
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 2.39,
|
| 258 |
+
"eval_accuracy": 0.9912970678711888,
|
| 259 |
+
"eval_f1": 0.9478974702630256,
|
| 260 |
+
"eval_loss": 0.03643050789833069,
|
| 261 |
+
"eval_precision": 0.9436290860573716,
|
| 262 |
+
"eval_recall": 0.9522046449007069,
|
| 263 |
+
"eval_runtime": 6.3303,
|
| 264 |
+
"eval_samples_per_second": 513.56,
|
| 265 |
+
"eval_steps_per_second": 0.632,
|
| 266 |
+
"step": 525
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 2.5,
|
| 270 |
+
"eval_accuracy": 0.9916085822203186,
|
| 271 |
+
"eval_f1": 0.9496052410549304,
|
| 272 |
+
"eval_loss": 0.03402611240744591,
|
| 273 |
+
"eval_precision": 0.9478537894030852,
|
| 274 |
+
"eval_recall": 0.9513631773813531,
|
| 275 |
+
"eval_runtime": 6.2277,
|
| 276 |
+
"eval_samples_per_second": 522.02,
|
| 277 |
+
"eval_steps_per_second": 0.642,
|
| 278 |
+
"step": 550
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"epoch": 2.61,
|
| 282 |
+
"eval_accuracy": 0.9917254001012422,
|
| 283 |
+
"eval_f1": 0.9519818989357244,
|
| 284 |
+
"eval_loss": 0.03392445296049118,
|
| 285 |
+
"eval_precision": 0.9480887998664663,
|
| 286 |
+
"eval_recall": 0.9559071019858634,
|
| 287 |
+
"eval_runtime": 6.3207,
|
| 288 |
+
"eval_samples_per_second": 514.343,
|
| 289 |
+
"eval_steps_per_second": 0.633,
|
| 290 |
+
"step": 575
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 2.73,
|
| 294 |
+
"eval_accuracy": 0.9901872980024142,
|
| 295 |
+
"eval_f1": 0.9414020171709594,
|
| 296 |
+
"eval_loss": 0.039618950337171555,
|
| 297 |
+
"eval_precision": 0.932617671345995,
|
| 298 |
+
"eval_recall": 0.9503534163581285,
|
| 299 |
+
"eval_runtime": 6.1153,
|
| 300 |
+
"eval_samples_per_second": 531.622,
|
| 301 |
+
"eval_steps_per_second": 0.654,
|
| 302 |
+
"step": 600
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 2.84,
|
| 306 |
+
"eval_accuracy": 0.9914722946925743,
|
| 307 |
+
"eval_f1": 0.9502345844504022,
|
| 308 |
+
"eval_loss": 0.03481649234890938,
|
| 309 |
+
"eval_precision": 0.9461127794461128,
|
| 310 |
+
"eval_recall": 0.9543924604510265,
|
| 311 |
+
"eval_runtime": 6.3663,
|
| 312 |
+
"eval_samples_per_second": 510.659,
|
| 313 |
+
"eval_steps_per_second": 0.628,
|
| 314 |
+
"step": 625
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.95,
|
| 318 |
+
"eval_accuracy": 0.9907908570538531,
|
| 319 |
+
"eval_f1": 0.9472891566265059,
|
| 320 |
+
"eval_loss": 0.03590630739927292,
|
| 321 |
+
"eval_precision": 0.9419301164725458,
|
| 322 |
+
"eval_recall": 0.9527095254123191,
|
| 323 |
+
"eval_runtime": 6.1326,
|
| 324 |
+
"eval_samples_per_second": 530.121,
|
| 325 |
+
"eval_steps_per_second": 0.652,
|
| 326 |
+
"step": 650
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 3.07,
|
| 330 |
+
"eval_accuracy": 0.9916475215139597,
|
| 331 |
+
"eval_f1": 0.9502965499958232,
|
| 332 |
+
"eval_loss": 0.03470408916473389,
|
| 333 |
+
"eval_precision": 0.9434400398075966,
|
| 334 |
+
"eval_recall": 0.9572534500168294,
|
| 335 |
+
"eval_runtime": 6.3596,
|
| 336 |
+
"eval_samples_per_second": 511.198,
|
| 337 |
+
"eval_steps_per_second": 0.629,
|
| 338 |
+
"step": 675
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 3.18,
|
| 342 |
+
"eval_accuracy": 0.9918227483353452,
|
| 343 |
+
"eval_f1": 0.9514563106796117,
|
| 344 |
+
"eval_loss": 0.03510947525501251,
|
| 345 |
+
"eval_precision": 0.9463869463869464,
|
| 346 |
+
"eval_recall": 0.9565802760013463,
|
| 347 |
+
"eval_runtime": 6.0809,
|
| 348 |
+
"eval_samples_per_second": 534.622,
|
| 349 |
+
"eval_steps_per_second": 0.658,
|
| 350 |
+
"step": 700
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 3.3,
|
| 354 |
+
"eval_accuracy": 0.9910634321093416,
|
| 355 |
+
"eval_f1": 0.9490787269681741,
|
| 356 |
+
"eval_loss": 0.03696366772055626,
|
| 357 |
+
"eval_precision": 0.944648216072024,
|
| 358 |
+
"eval_recall": 0.9535509929316729,
|
| 359 |
+
"eval_runtime": 6.1802,
|
| 360 |
+
"eval_samples_per_second": 526.033,
|
| 361 |
+
"eval_steps_per_second": 0.647,
|
| 362 |
+
"step": 725
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 3.41,
|
| 366 |
+
"eval_accuracy": 0.9916669911607804,
|
| 367 |
+
"eval_f1": 0.9521739130434784,
|
| 368 |
+
"eval_loss": 0.03580446541309357,
|
| 369 |
+
"eval_precision": 0.9461615154536391,
|
| 370 |
+
"eval_recall": 0.9582632110400539,
|
| 371 |
+
"eval_runtime": 6.3604,
|
| 372 |
+
"eval_samples_per_second": 511.128,
|
| 373 |
+
"eval_steps_per_second": 0.629,
|
| 374 |
+
"step": 750
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 3.52,
|
| 378 |
+
"eval_accuracy": 0.9919785055099101,
|
| 379 |
+
"eval_f1": 0.9523250942605781,
|
| 380 |
+
"eval_loss": 0.03529668599367142,
|
| 381 |
+
"eval_precision": 0.9482729851493409,
|
| 382 |
+
"eval_recall": 0.9564119824974756,
|
| 383 |
+
"eval_runtime": 6.1111,
|
| 384 |
+
"eval_samples_per_second": 531.982,
|
| 385 |
+
"eval_steps_per_second": 0.655,
|
| 386 |
+
"step": 775
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.64,
|
| 390 |
+
"eval_accuracy": 0.9915696429266773,
|
| 391 |
+
"eval_f1": 0.9516075016744809,
|
| 392 |
+
"eval_loss": 0.035107776522636414,
|
| 393 |
+
"eval_precision": 0.9468510496501167,
|
| 394 |
+
"eval_recall": 0.9564119824974756,
|
| 395 |
+
"eval_runtime": 6.1664,
|
| 396 |
+
"eval_samples_per_second": 527.215,
|
| 397 |
+
"eval_steps_per_second": 0.649,
|
| 398 |
+
"step": 800
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 3.75,
|
| 402 |
+
"eval_accuracy": 0.9918811572758071,
|
| 403 |
+
"eval_f1": 0.9528751987946764,
|
| 404 |
+
"eval_loss": 0.036074355244636536,
|
| 405 |
+
"eval_precision": 0.9478767693588677,
|
| 406 |
+
"eval_recall": 0.9579266240323123,
|
| 407 |
+
"eval_runtime": 6.3953,
|
| 408 |
+
"eval_samples_per_second": 508.343,
|
| 409 |
+
"eval_steps_per_second": 0.625,
|
| 410 |
+
"step": 825
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 3.86,
|
| 414 |
+
"eval_accuracy": 0.9918032786885246,
|
| 415 |
+
"eval_f1": 0.9539209115281501,
|
| 416 |
+
"eval_loss": 0.0369994230568409,
|
| 417 |
+
"eval_precision": 0.9497831164497831,
|
| 418 |
+
"eval_recall": 0.9580949175361831,
|
| 419 |
+
"eval_runtime": 6.198,
|
| 420 |
+
"eval_samples_per_second": 524.528,
|
| 421 |
+
"eval_steps_per_second": 0.645,
|
| 422 |
+
"step": 850
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 3.98,
|
| 426 |
+
"eval_accuracy": 0.9915112339862155,
|
| 427 |
+
"eval_f1": 0.9516560722649716,
|
| 428 |
+
"eval_loss": 0.03739459067583084,
|
| 429 |
+
"eval_precision": 0.9459594280013303,
|
| 430 |
+
"eval_recall": 0.9574217435207001,
|
| 431 |
+
"eval_runtime": 6.4107,
|
| 432 |
+
"eval_samples_per_second": 507.122,
|
| 433 |
+
"eval_steps_per_second": 0.624,
|
| 434 |
+
"step": 875
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"epoch": 4.09,
|
| 438 |
+
"eval_accuracy": 0.9921732019781161,
|
| 439 |
+
"eval_f1": 0.9550213585727447,
|
| 440 |
+
"eval_loss": 0.03806319832801819,
|
| 441 |
+
"eval_precision": 0.9506419876604969,
|
| 442 |
+
"eval_recall": 0.9594412655671491,
|
| 443 |
+
"eval_runtime": 6.1503,
|
| 444 |
+
"eval_samples_per_second": 528.594,
|
| 445 |
+
"eval_steps_per_second": 0.65,
|
| 446 |
+
"step": 900
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 4.2,
|
| 450 |
+
"eval_accuracy": 0.9911997196370858,
|
| 451 |
+
"eval_f1": 0.9508580996232734,
|
| 452 |
+
"eval_loss": 0.041537731885910034,
|
| 453 |
+
"eval_precision": 0.9460269865067467,
|
| 454 |
+
"eval_recall": 0.9557388084819926,
|
| 455 |
+
"eval_runtime": 6.222,
|
| 456 |
+
"eval_samples_per_second": 522.501,
|
| 457 |
+
"eval_steps_per_second": 0.643,
|
| 458 |
+
"step": 925
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 4.32,
|
| 462 |
+
"eval_accuracy": 0.9916864608076009,
|
| 463 |
+
"eval_f1": 0.9524448544829321,
|
| 464 |
+
"eval_loss": 0.03895916789770126,
|
| 465 |
+
"eval_precision": 0.9493395753218525,
|
| 466 |
+
"eval_recall": 0.9555705149781218,
|
| 467 |
+
"eval_runtime": 6.4146,
|
| 468 |
+
"eval_samples_per_second": 506.815,
|
| 469 |
+
"eval_steps_per_second": 0.624,
|
| 470 |
+
"step": 950
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 4.43,
|
| 474 |
+
"eval_accuracy": 0.9919200965694482,
|
| 475 |
+
"eval_f1": 0.9536479250334672,
|
| 476 |
+
"eval_loss": 0.038863833993673325,
|
| 477 |
+
"eval_precision": 0.9482529118136439,
|
| 478 |
+
"eval_recall": 0.9591046785594076,
|
| 479 |
+
"eval_runtime": 6.2052,
|
| 480 |
+
"eval_samples_per_second": 523.911,
|
| 481 |
+
"eval_steps_per_second": 0.645,
|
| 482 |
+
"step": 975
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 4.55,
|
| 486 |
+
"learning_rate": 2.7272727272727273e-05,
|
| 487 |
+
"loss": 0.0123,
|
| 488 |
+
"step": 1000
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 4.55,
|
| 492 |
+
"eval_accuracy": 0.9918227483353452,
|
| 493 |
+
"eval_f1": 0.95163179916318,
|
| 494 |
+
"eval_loss": 0.03789864480495453,
|
| 495 |
+
"eval_precision": 0.9464047936085219,
|
| 496 |
+
"eval_recall": 0.9569168630090878,
|
| 497 |
+
"eval_runtime": 6.4251,
|
| 498 |
+
"eval_samples_per_second": 505.987,
|
| 499 |
+
"eval_steps_per_second": 0.623,
|
| 500 |
+
"step": 1000
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 4.66,
|
| 504 |
+
"eval_accuracy": 0.9919785055099101,
|
| 505 |
+
"eval_f1": 0.9520782805051433,
|
| 506 |
+
"eval_loss": 0.037610627710819244,
|
| 507 |
+
"eval_precision": 0.9463009143807148,
|
| 508 |
+
"eval_recall": 0.9579266240323123,
|
| 509 |
+
"eval_runtime": 5.9694,
|
| 510 |
+
"eval_samples_per_second": 544.608,
|
| 511 |
+
"eval_steps_per_second": 0.67,
|
| 512 |
+
"step": 1025
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 4.77,
|
| 516 |
+
"eval_accuracy": 0.9917059304544216,
|
| 517 |
+
"eval_f1": 0.9534747254589656,
|
| 518 |
+
"eval_loss": 0.03730636462569237,
|
| 519 |
+
"eval_precision": 0.9498914314347754,
|
| 520 |
+
"eval_recall": 0.9570851565129586,
|
| 521 |
+
"eval_runtime": 6.4491,
|
| 522 |
+
"eval_samples_per_second": 504.105,
|
| 523 |
+
"eval_steps_per_second": 0.62,
|
| 524 |
+
"step": 1050
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 4.89,
|
| 528 |
+
"eval_accuracy": 0.9922510805653986,
|
| 529 |
+
"eval_f1": 0.9552163703455216,
|
| 530 |
+
"eval_loss": 0.03659652918577194,
|
| 531 |
+
"eval_precision": 0.9520227348712805,
|
| 532 |
+
"eval_recall": 0.9584315045439246,
|
| 533 |
+
"eval_runtime": 6.1088,
|
| 534 |
+
"eval_samples_per_second": 532.187,
|
| 535 |
+
"eval_steps_per_second": 0.655,
|
| 536 |
+
"step": 1075
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 5.0,
|
| 540 |
+
"eval_accuracy": 0.9922705502122191,
|
| 541 |
+
"eval_f1": 0.9546746947650109,
|
| 542 |
+
"eval_loss": 0.03736859932541847,
|
| 543 |
+
"eval_precision": 0.9488031914893617,
|
| 544 |
+
"eval_recall": 0.9606193200942443,
|
| 545 |
+
"eval_runtime": 6.1341,
|
| 546 |
+
"eval_samples_per_second": 529.989,
|
| 547 |
+
"eval_steps_per_second": 0.652,
|
| 548 |
+
"step": 1100
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 5.11,
|
| 552 |
+
"eval_accuracy": 0.9920369144503719,
|
| 553 |
+
"eval_f1": 0.9552388935456831,
|
| 554 |
+
"eval_loss": 0.039282701909542084,
|
| 555 |
+
"eval_precision": 0.9515698062792252,
|
| 556 |
+
"eval_recall": 0.9589363850555369,
|
| 557 |
+
"eval_runtime": 6.409,
|
| 558 |
+
"eval_samples_per_second": 507.255,
|
| 559 |
+
"eval_steps_per_second": 0.624,
|
| 560 |
+
"step": 1125
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 5.23,
|
| 564 |
+
"eval_accuracy": 0.9925431252677076,
|
| 565 |
+
"eval_f1": 0.9570613887957062,
|
| 566 |
+
"eval_loss": 0.038866784423589706,
|
| 567 |
+
"eval_precision": 0.9538615847542627,
|
| 568 |
+
"eval_recall": 0.9602827330865029,
|
| 569 |
+
"eval_runtime": 6.1822,
|
| 570 |
+
"eval_samples_per_second": 525.868,
|
| 571 |
+
"eval_steps_per_second": 0.647,
|
| 572 |
+
"step": 1150
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 5.34,
|
| 576 |
+
"eval_accuracy": 0.9916864608076009,
|
| 577 |
+
"eval_f1": 0.9530988274706869,
|
| 578 |
+
"eval_loss": 0.039710164070129395,
|
| 579 |
+
"eval_precision": 0.94864954984995,
|
| 580 |
+
"eval_recall": 0.9575900370245709,
|
| 581 |
+
"eval_runtime": 6.3284,
|
| 582 |
+
"eval_samples_per_second": 513.715,
|
| 583 |
+
"eval_steps_per_second": 0.632,
|
| 584 |
+
"step": 1175
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 5.45,
|
| 588 |
+
"eval_accuracy": 0.9919395662162689,
|
| 589 |
+
"eval_f1": 0.9523490494933422,
|
| 590 |
+
"eval_loss": 0.03970210626721382,
|
| 591 |
+
"eval_precision": 0.9478246374395732,
|
| 592 |
+
"eval_recall": 0.9569168630090878,
|
| 593 |
+
"eval_runtime": 6.2875,
|
| 594 |
+
"eval_samples_per_second": 517.058,
|
| 595 |
+
"eval_steps_per_second": 0.636,
|
| 596 |
+
"step": 1200
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 5.57,
|
| 600 |
+
"eval_accuracy": 0.9920174448035513,
|
| 601 |
+
"eval_f1": 0.9537354639002761,
|
| 602 |
+
"eval_loss": 0.03881015256047249,
|
| 603 |
+
"eval_precision": 0.9482615205456663,
|
| 604 |
+
"eval_recall": 0.9592729720632783,
|
| 605 |
+
"eval_runtime": 6.0937,
|
| 606 |
+
"eval_samples_per_second": 533.498,
|
| 607 |
+
"eval_steps_per_second": 0.656,
|
| 608 |
+
"step": 1225
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 5.68,
|
| 612 |
+
"eval_accuracy": 0.9922900198590398,
|
| 613 |
+
"eval_f1": 0.9553937567997322,
|
| 614 |
+
"eval_loss": 0.038882434368133545,
|
| 615 |
+
"eval_precision": 0.9502247378058931,
|
| 616 |
+
"eval_recall": 0.9606193200942443,
|
| 617 |
+
"eval_runtime": 6.4419,
|
| 618 |
+
"eval_samples_per_second": 504.662,
|
| 619 |
+
"eval_steps_per_second": 0.621,
|
| 620 |
+
"step": 1250
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 5.8,
|
| 624 |
+
"eval_accuracy": 0.9925431252677076,
|
| 625 |
+
"eval_f1": 0.958162153098013,
|
| 626 |
+
"eval_loss": 0.038017645478248596,
|
| 627 |
+
"eval_precision": 0.9547201336675021,
|
| 628 |
+
"eval_recall": 0.9616290811174689,
|
| 629 |
+
"eval_runtime": 6.1177,
|
| 630 |
+
"eval_samples_per_second": 531.411,
|
| 631 |
+
"eval_steps_per_second": 0.654,
|
| 632 |
+
"step": 1275
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 5.91,
|
| 636 |
+
"eval_accuracy": 0.9924263073867841,
|
| 637 |
+
"eval_f1": 0.954899171617438,
|
| 638 |
+
"eval_loss": 0.03911532089114189,
|
| 639 |
+
"eval_precision": 0.9495756365451822,
|
| 640 |
+
"eval_recall": 0.9602827330865029,
|
| 641 |
+
"eval_runtime": 6.3742,
|
| 642 |
+
"eval_samples_per_second": 510.022,
|
| 643 |
+
"eval_steps_per_second": 0.628,
|
| 644 |
+
"step": 1300
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 6.02,
|
| 648 |
+
"eval_accuracy": 0.9924457770336046,
|
| 649 |
+
"eval_f1": 0.9575432119483135,
|
| 650 |
+
"eval_loss": 0.038087744265794754,
|
| 651 |
+
"eval_precision": 0.9548192771084337,
|
| 652 |
+
"eval_recall": 0.9602827330865029,
|
| 653 |
+
"eval_runtime": 6.0894,
|
| 654 |
+
"eval_samples_per_second": 533.875,
|
| 655 |
+
"eval_steps_per_second": 0.657,
|
| 656 |
+
"step": 1325
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 6.14,
|
| 660 |
+
"eval_accuracy": 0.9921926716249367,
|
| 661 |
+
"eval_f1": 0.9562300855274191,
|
| 662 |
+
"eval_loss": 0.03999880701303482,
|
| 663 |
+
"eval_precision": 0.9528743315508021,
|
| 664 |
+
"eval_recall": 0.9596095590710199,
|
| 665 |
+
"eval_runtime": 6.1345,
|
| 666 |
+
"eval_samples_per_second": 529.956,
|
| 667 |
+
"eval_steps_per_second": 0.652,
|
| 668 |
+
"step": 1350
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 6.25,
|
| 672 |
+
"eval_accuracy": 0.9927378217359137,
|
| 673 |
+
"eval_f1": 0.958001508927823,
|
| 674 |
+
"eval_loss": 0.03927755728363991,
|
| 675 |
+
"eval_precision": 0.9544012026056455,
|
| 676 |
+
"eval_recall": 0.9616290811174689,
|
| 677 |
+
"eval_runtime": 6.3512,
|
| 678 |
+
"eval_samples_per_second": 511.868,
|
| 679 |
+
"eval_steps_per_second": 0.63,
|
| 680 |
+
"step": 1375
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 6.36,
|
| 684 |
+
"eval_accuracy": 0.9924263073867841,
|
| 685 |
+
"eval_f1": 0.9567400217555017,
|
| 686 |
+
"eval_loss": 0.04191247373819351,
|
| 687 |
+
"eval_precision": 0.9514062239973373,
|
| 688 |
+
"eval_recall": 0.9621339616290812,
|
| 689 |
+
"eval_runtime": 6.1376,
|
| 690 |
+
"eval_samples_per_second": 529.682,
|
| 691 |
+
"eval_steps_per_second": 0.652,
|
| 692 |
+
"step": 1400
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 6.48,
|
| 696 |
+
"eval_accuracy": 0.9925431252677076,
|
| 697 |
+
"eval_f1": 0.9578832789081471,
|
| 698 |
+
"eval_loss": 0.04148797690868378,
|
| 699 |
+
"eval_precision": 0.9531744709215131,
|
| 700 |
+
"eval_recall": 0.9626388421406934,
|
| 701 |
+
"eval_runtime": 6.379,
|
| 702 |
+
"eval_samples_per_second": 509.643,
|
| 703 |
+
"eval_steps_per_second": 0.627,
|
| 704 |
+
"step": 1425
|
| 705 |
+
},
|
| 706 |
+
{
|
| 707 |
+
"epoch": 6.59,
|
| 708 |
+
"eval_accuracy": 0.9923289591526809,
|
| 709 |
+
"eval_f1": 0.9566236811254396,
|
| 710 |
+
"eval_loss": 0.04145730286836624,
|
| 711 |
+
"eval_precision": 0.952,
|
| 712 |
+
"eval_recall": 0.9612924941097274,
|
| 713 |
+
"eval_runtime": 6.1372,
|
| 714 |
+
"eval_samples_per_second": 529.721,
|
| 715 |
+
"eval_steps_per_second": 0.652,
|
| 716 |
+
"step": 1450
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 6.7,
|
| 720 |
+
"eval_accuracy": 0.9925236556208871,
|
| 721 |
+
"eval_f1": 0.9576590928146222,
|
| 722 |
+
"eval_loss": 0.03992067277431488,
|
| 723 |
+
"eval_precision": 0.95421888053467,
|
| 724 |
+
"eval_recall": 0.9611242006058566,
|
| 725 |
+
"eval_runtime": 6.4046,
|
| 726 |
+
"eval_samples_per_second": 507.601,
|
| 727 |
+
"eval_steps_per_second": 0.625,
|
| 728 |
+
"step": 1475
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 6.82,
|
| 732 |
+
"learning_rate": 1.590909090909091e-05,
|
| 733 |
+
"loss": 0.0052,
|
| 734 |
+
"step": 1500
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 6.82,
|
| 738 |
+
"eval_accuracy": 0.9920563840971924,
|
| 739 |
+
"eval_f1": 0.9556468516810598,
|
| 740 |
+
"eval_loss": 0.0415743812918663,
|
| 741 |
+
"eval_precision": 0.9522138680033417,
|
| 742 |
+
"eval_recall": 0.9591046785594076,
|
| 743 |
+
"eval_runtime": 6.2338,
|
| 744 |
+
"eval_samples_per_second": 521.515,
|
| 745 |
+
"eval_steps_per_second": 0.642,
|
| 746 |
+
"step": 1500
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"epoch": 6.93,
|
| 750 |
+
"eval_accuracy": 0.9919006269226276,
|
| 751 |
+
"eval_f1": 0.9550439514441188,
|
| 752 |
+
"eval_loss": 0.04101209715008736,
|
| 753 |
+
"eval_precision": 0.9501915708812261,
|
| 754 |
+
"eval_recall": 0.9599461460787614,
|
| 755 |
+
"eval_runtime": 5.9759,
|
| 756 |
+
"eval_samples_per_second": 544.016,
|
| 757 |
+
"eval_steps_per_second": 0.669,
|
| 758 |
+
"step": 1525
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 7.05,
|
| 762 |
+
"eval_accuracy": 0.9920758537440131,
|
| 763 |
+
"eval_f1": 0.9559832635983264,
|
| 764 |
+
"eval_loss": 0.040564827620983124,
|
| 765 |
+
"eval_precision": 0.9507323568575233,
|
| 766 |
+
"eval_recall": 0.9612924941097274,
|
| 767 |
+
"eval_runtime": 6.3076,
|
| 768 |
+
"eval_samples_per_second": 515.41,
|
| 769 |
+
"eval_steps_per_second": 0.634,
|
| 770 |
+
"step": 1550
|
| 771 |
+
},
|
| 772 |
+
{
|
| 773 |
+
"epoch": 7.16,
|
| 774 |
+
"eval_accuracy": 0.9922705502122191,
|
| 775 |
+
"eval_f1": 0.9555388093443858,
|
| 776 |
+
"eval_loss": 0.039956897497177124,
|
| 777 |
+
"eval_precision": 0.9508415264122646,
|
| 778 |
+
"eval_recall": 0.9602827330865029,
|
| 779 |
+
"eval_runtime": 6.1016,
|
| 780 |
+
"eval_samples_per_second": 532.811,
|
| 781 |
+
"eval_steps_per_second": 0.656,
|
| 782 |
+
"step": 1575
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 7.27,
|
| 786 |
+
"eval_accuracy": 0.9924263073867841,
|
| 787 |
+
"eval_f1": 0.957126109529392,
|
| 788 |
+
"eval_loss": 0.04024273902177811,
|
| 789 |
+
"eval_precision": 0.9525,
|
| 790 |
+
"eval_recall": 0.9617973746213396,
|
| 791 |
+
"eval_runtime": 6.1018,
|
| 792 |
+
"eval_samples_per_second": 532.791,
|
| 793 |
+
"eval_steps_per_second": 0.656,
|
| 794 |
+
"step": 1600
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 7.39,
|
| 798 |
+
"eval_accuracy": 0.9924847163272458,
|
| 799 |
+
"eval_f1": 0.9591152815013404,
|
| 800 |
+
"eval_loss": 0.04005241021513939,
|
| 801 |
+
"eval_precision": 0.954954954954955,
|
| 802 |
+
"eval_recall": 0.9633120161561763,
|
| 803 |
+
"eval_runtime": 6.4344,
|
| 804 |
+
"eval_samples_per_second": 505.253,
|
| 805 |
+
"eval_steps_per_second": 0.622,
|
| 806 |
+
"step": 1625
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 7.5,
|
| 810 |
+
"eval_accuracy": 0.9927183520890931,
|
| 811 |
+
"eval_f1": 0.9600535968511849,
|
| 812 |
+
"eval_loss": 0.039704449474811554,
|
| 813 |
+
"eval_precision": 0.9554925820970162,
|
| 814 |
+
"eval_recall": 0.9646583641871423,
|
| 815 |
+
"eval_runtime": 6.185,
|
| 816 |
+
"eval_samples_per_second": 525.625,
|
| 817 |
+
"eval_steps_per_second": 0.647,
|
| 818 |
+
"step": 1650
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 7.61,
|
| 822 |
+
"eval_accuracy": 0.9921732019781161,
|
| 823 |
+
"eval_f1": 0.956769436997319,
|
| 824 |
+
"eval_loss": 0.04118635505437851,
|
| 825 |
+
"eval_precision": 0.9526192859526192,
|
| 826 |
+
"eval_recall": 0.9609559071019859,
|
| 827 |
+
"eval_runtime": 6.2989,
|
| 828 |
+
"eval_samples_per_second": 516.125,
|
| 829 |
+
"eval_steps_per_second": 0.635,
|
| 830 |
+
"step": 1675
|
| 831 |
+
},
|
| 832 |
+
{
|
| 833 |
+
"epoch": 7.73,
|
| 834 |
+
"eval_accuracy": 0.9923289591526809,
|
| 835 |
+
"eval_f1": 0.9573594705537405,
|
| 836 |
+
"eval_loss": 0.041853681206703186,
|
| 837 |
+
"eval_precision": 0.9531276063386155,
|
| 838 |
+
"eval_recall": 0.9616290811174689,
|
| 839 |
+
"eval_runtime": 6.0836,
|
| 840 |
+
"eval_samples_per_second": 534.391,
|
| 841 |
+
"eval_steps_per_second": 0.658,
|
| 842 |
+
"step": 1700
|
| 843 |
+
},
|
| 844 |
+
{
|
| 845 |
+
"epoch": 7.84,
|
| 846 |
+
"eval_accuracy": 0.9927378217359137,
|
| 847 |
+
"eval_f1": 0.9588259958071279,
|
| 848 |
+
"eval_loss": 0.040749866515398026,
|
| 849 |
+
"eval_precision": 0.9555406986461641,
|
| 850 |
+
"eval_recall": 0.9621339616290812,
|
| 851 |
+
"eval_runtime": 6.3737,
|
| 852 |
+
"eval_samples_per_second": 510.067,
|
| 853 |
+
"eval_steps_per_second": 0.628,
|
| 854 |
+
"step": 1725
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 7.95,
|
| 858 |
+
"eval_accuracy": 0.9926988824422724,
|
| 859 |
+
"eval_f1": 0.9589339591015756,
|
| 860 |
+
"eval_loss": 0.04090812802314758,
|
| 861 |
+
"eval_precision": 0.9550918196994992,
|
| 862 |
+
"eval_recall": 0.9628071356445641,
|
| 863 |
+
"eval_runtime": 6.1325,
|
| 864 |
+
"eval_samples_per_second": 530.13,
|
| 865 |
+
"eval_steps_per_second": 0.652,
|
| 866 |
+
"step": 1750
|
| 867 |
+
},
|
| 868 |
+
{
|
| 869 |
+
"epoch": 8.07,
|
| 870 |
+
"eval_accuracy": 0.9923873680931428,
|
| 871 |
+
"eval_f1": 0.9567983924983255,
|
| 872 |
+
"eval_loss": 0.041253186762332916,
|
| 873 |
+
"eval_precision": 0.9520159946684439,
|
| 874 |
+
"eval_recall": 0.9616290811174689,
|
| 875 |
+
"eval_runtime": 6.1524,
|
| 876 |
+
"eval_samples_per_second": 528.415,
|
| 877 |
+
"eval_steps_per_second": 0.65,
|
| 878 |
+
"step": 1775
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 8.18,
|
| 882 |
+
"eval_accuracy": 0.9922510805653986,
|
| 883 |
+
"eval_f1": 0.95546626485853,
|
| 884 |
+
"eval_loss": 0.041376810520887375,
|
| 885 |
+
"eval_precision": 0.9505329780146569,
|
| 886 |
+
"eval_recall": 0.9604510265903736,
|
| 887 |
+
"eval_runtime": 6.3306,
|
| 888 |
+
"eval_samples_per_second": 513.537,
|
| 889 |
+
"eval_steps_per_second": 0.632,
|
| 890 |
+
"step": 1800
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 8.3,
|
| 894 |
+
"eval_accuracy": 0.9924263073867841,
|
| 895 |
+
"eval_f1": 0.957309401996142,
|
| 896 |
+
"eval_loss": 0.04096854478120804,
|
| 897 |
+
"eval_precision": 0.9541882628323023,
|
| 898 |
+
"eval_recall": 0.9604510265903736,
|
| 899 |
+
"eval_runtime": 6.092,
|
| 900 |
+
"eval_samples_per_second": 533.646,
|
| 901 |
+
"eval_steps_per_second": 0.657,
|
| 902 |
+
"step": 1825
|
| 903 |
+
},
|
| 904 |
+
{
|
| 905 |
+
"epoch": 8.41,
|
| 906 |
+
"eval_accuracy": 0.9924457770336046,
|
| 907 |
+
"eval_f1": 0.9576093343406362,
|
| 908 |
+
"eval_loss": 0.04169493913650513,
|
| 909 |
+
"eval_precision": 0.9552838720482332,
|
| 910 |
+
"eval_recall": 0.9599461460787614,
|
| 911 |
+
"eval_runtime": 6.3572,
|
| 912 |
+
"eval_samples_per_second": 511.386,
|
| 913 |
+
"eval_steps_per_second": 0.629,
|
| 914 |
+
"step": 1850
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 8.52,
|
| 918 |
+
"eval_accuracy": 0.9923289591526809,
|
| 919 |
+
"eval_f1": 0.9575574568025499,
|
| 920 |
+
"eval_loss": 0.04175976663827896,
|
| 921 |
+
"eval_precision": 0.954515050167224,
|
| 922 |
+
"eval_recall": 0.9606193200942443,
|
| 923 |
+
"eval_runtime": 6.2169,
|
| 924 |
+
"eval_samples_per_second": 522.929,
|
| 925 |
+
"eval_steps_per_second": 0.643,
|
| 926 |
+
"step": 1875
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 8.64,
|
| 930 |
+
"eval_accuracy": 0.9924263073867841,
|
| 931 |
+
"eval_f1": 0.958001508927823,
|
| 932 |
+
"eval_loss": 0.04138173907995224,
|
| 933 |
+
"eval_precision": 0.9544012026056455,
|
| 934 |
+
"eval_recall": 0.9616290811174689,
|
| 935 |
+
"eval_runtime": 6.4751,
|
| 936 |
+
"eval_samples_per_second": 502.077,
|
| 937 |
+
"eval_steps_per_second": 0.618,
|
| 938 |
+
"step": 1900
|
| 939 |
+
},
|
| 940 |
+
{
|
| 941 |
+
"epoch": 8.75,
|
| 942 |
+
"eval_accuracy": 0.9925236556208871,
|
| 943 |
+
"eval_f1": 0.9587386782958739,
|
| 944 |
+
"eval_loss": 0.04193359240889549,
|
| 945 |
+
"eval_precision": 0.9555332664660648,
|
| 946 |
+
"eval_recall": 0.9619656681252103,
|
| 947 |
+
"eval_runtime": 6.1398,
|
| 948 |
+
"eval_samples_per_second": 529.492,
|
| 949 |
+
"eval_steps_per_second": 0.651,
|
| 950 |
+
"step": 1925
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 8.86,
|
| 954 |
+
"eval_accuracy": 0.9925625949145283,
|
| 955 |
+
"eval_f1": 0.9577393929230252,
|
| 956 |
+
"eval_loss": 0.04151003435254097,
|
| 957 |
+
"eval_precision": 0.9543783422459893,
|
| 958 |
+
"eval_recall": 0.9611242006058566,
|
| 959 |
+
"eval_runtime": 6.1314,
|
| 960 |
+
"eval_samples_per_second": 530.223,
|
| 961 |
+
"eval_steps_per_second": 0.652,
|
| 962 |
+
"step": 1950
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 8.98,
|
| 966 |
+
"eval_accuracy": 0.9926210038549901,
|
| 967 |
+
"eval_f1": 0.9576590928146222,
|
| 968 |
+
"eval_loss": 0.041335199028253555,
|
| 969 |
+
"eval_precision": 0.95421888053467,
|
| 970 |
+
"eval_recall": 0.9611242006058566,
|
| 971 |
+
"eval_runtime": 6.4106,
|
| 972 |
+
"eval_samples_per_second": 507.128,
|
| 973 |
+
"eval_steps_per_second": 0.624,
|
| 974 |
+
"step": 1975
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 9.09,
|
| 978 |
+
"learning_rate": 4.5454545454545455e-06,
|
| 979 |
+
"loss": 0.0027,
|
| 980 |
+
"step": 2000
|
| 981 |
+
},
|
| 982 |
+
{
|
| 983 |
+
"epoch": 9.09,
|
| 984 |
+
"eval_accuracy": 0.9926599431486313,
|
| 985 |
+
"eval_f1": 0.9590143324113654,
|
| 986 |
+
"eval_loss": 0.0412251353263855,
|
| 987 |
+
"eval_precision": 0.9552512940390716,
|
| 988 |
+
"eval_recall": 0.9628071356445641,
|
| 989 |
+
"eval_runtime": 6.2732,
|
| 990 |
+
"eval_samples_per_second": 518.239,
|
| 991 |
+
"eval_steps_per_second": 0.638,
|
| 992 |
+
"step": 2000
|
| 993 |
+
},
|
| 994 |
+
{
|
| 995 |
+
"epoch": 9.2,
|
| 996 |
+
"eval_accuracy": 0.9926599431486313,
|
| 997 |
+
"eval_f1": 0.9591819629536502,
|
| 998 |
+
"eval_loss": 0.04075352102518082,
|
| 999 |
+
"eval_precision": 0.955418266822508,
|
| 1000 |
+
"eval_recall": 0.9629754291484349,
|
| 1001 |
+
"eval_runtime": 6.2741,
|
| 1002 |
+
"eval_samples_per_second": 518.162,
|
| 1003 |
+
"eval_steps_per_second": 0.638,
|
| 1004 |
+
"step": 2025
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 9.32,
|
| 1008 |
+
"eval_accuracy": 0.9926210038549901,
|
| 1009 |
+
"eval_f1": 0.9579070937447594,
|
| 1010 |
+
"eval_loss": 0.040350958704948425,
|
| 1011 |
+
"eval_precision": 0.9545454545454546,
|
| 1012 |
+
"eval_recall": 0.9612924941097274,
|
| 1013 |
+
"eval_runtime": 6.0983,
|
| 1014 |
+
"eval_samples_per_second": 533.103,
|
| 1015 |
+
"eval_steps_per_second": 0.656,
|
| 1016 |
+
"step": 2050
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 9.43,
|
| 1020 |
+
"eval_accuracy": 0.9926404735018107,
|
| 1021 |
+
"eval_f1": 0.9587317564167085,
|
| 1022 |
+
"eval_loss": 0.040736954659223557,
|
| 1023 |
+
"eval_precision": 0.955685618729097,
|
| 1024 |
+
"eval_recall": 0.9617973746213396,
|
| 1025 |
+
"eval_runtime": 6.0517,
|
| 1026 |
+
"eval_samples_per_second": 537.203,
|
| 1027 |
+
"eval_steps_per_second": 0.661,
|
| 1028 |
+
"step": 2075
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 9.55,
|
| 1032 |
+
"eval_accuracy": 0.9926015342081694,
|
| 1033 |
+
"eval_f1": 0.9584905660377357,
|
| 1034 |
+
"eval_loss": 0.04098973423242569,
|
| 1035 |
+
"eval_precision": 0.9552064181848571,
|
| 1036 |
+
"eval_recall": 0.9617973746213396,
|
| 1037 |
+
"eval_runtime": 6.3854,
|
| 1038 |
+
"eval_samples_per_second": 509.128,
|
| 1039 |
+
"eval_steps_per_second": 0.626,
|
| 1040 |
+
"step": 2100
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 9.66,
|
| 1044 |
+
"eval_accuracy": 0.9925041859740664,
|
| 1045 |
+
"eval_f1": 0.9585778970316955,
|
| 1046 |
+
"eval_loss": 0.04119739308953285,
|
| 1047 |
+
"eval_precision": 0.9552139037433155,
|
| 1048 |
+
"eval_recall": 0.9619656681252103,
|
| 1049 |
+
"eval_runtime": 6.1845,
|
| 1050 |
+
"eval_samples_per_second": 525.672,
|
| 1051 |
+
"eval_steps_per_second": 0.647,
|
| 1052 |
+
"step": 2125
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 9.77,
|
| 1056 |
+
"eval_accuracy": 0.9924847163272458,
|
| 1057 |
+
"eval_f1": 0.9589064072458906,
|
| 1058 |
+
"eval_loss": 0.04128846526145935,
|
| 1059 |
+
"eval_precision": 0.9557004346372451,
|
| 1060 |
+
"eval_recall": 0.9621339616290812,
|
| 1061 |
+
"eval_runtime": 6.444,
|
| 1062 |
+
"eval_samples_per_second": 504.5,
|
| 1063 |
+
"eval_steps_per_second": 0.621,
|
| 1064 |
+
"step": 2150
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 9.89,
|
| 1068 |
+
"eval_accuracy": 0.9925041859740664,
|
| 1069 |
+
"eval_f1": 0.9589064072458906,
|
| 1070 |
+
"eval_loss": 0.04132496938109398,
|
| 1071 |
+
"eval_precision": 0.9557004346372451,
|
| 1072 |
+
"eval_recall": 0.9621339616290812,
|
| 1073 |
+
"eval_runtime": 6.0985,
|
| 1074 |
+
"eval_samples_per_second": 533.086,
|
| 1075 |
+
"eval_steps_per_second": 0.656,
|
| 1076 |
+
"step": 2175
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 10.0,
|
| 1080 |
+
"eval_accuracy": 0.9925041859740664,
|
| 1081 |
+
"eval_f1": 0.9589064072458906,
|
| 1082 |
+
"eval_loss": 0.041328806430101395,
|
| 1083 |
+
"eval_precision": 0.9557004346372451,
|
| 1084 |
+
"eval_recall": 0.9621339616290812,
|
| 1085 |
+
"eval_runtime": 6.1664,
|
| 1086 |
+
"eval_samples_per_second": 527.211,
|
| 1087 |
+
"eval_steps_per_second": 0.649,
|
| 1088 |
+
"step": 2200
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 10.0,
|
| 1092 |
+
"step": 2200,
|
| 1093 |
+
"total_flos": 4144211411102244.0,
|
| 1094 |
+
"train_loss": 0.022697925716638565,
|
| 1095 |
+
"train_runtime": 1311.4885,
|
| 1096 |
+
"train_samples_per_second": 107.069,
|
| 1097 |
+
"train_steps_per_second": 1.677
|
| 1098 |
+
}
|
| 1099 |
+
],
|
| 1100 |
+
"max_steps": 2200,
|
| 1101 |
+
"num_train_epochs": 10,
|
| 1102 |
+
"total_flos": 4144211411102244.0,
|
| 1103 |
+
"trial_name": null,
|
| 1104 |
+
"trial_params": null
|
| 1105 |
+
}
|