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": 0.9925820645613489,
|
| 4 |
+
"eval_f1": 0.9571847507331379,
|
| 5 |
+
"eval_loss": 0.04227477312088013,
|
| 6 |
+
"eval_precision": 0.953111963957951,
|
| 7 |
+
"eval_recall": 0.9612924941097274,
|
| 8 |
+
"eval_runtime": 6.2758,
|
| 9 |
+
"eval_samples": 3251,
|
| 10 |
+
"eval_samples_per_second": 518.019,
|
| 11 |
+
"eval_steps_per_second": 0.637,
|
| 12 |
+
"train_loss": 0.021953743980689482,
|
| 13 |
+
"train_runtime": 1347.1839,
|
| 14 |
+
"train_samples": 14042,
|
| 15 |
+
"train_samples_per_second": 104.232,
|
| 16 |
+
"train_steps_per_second": 1.633
|
| 17 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,12 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_accuracy": 0.9925820645613489,
|
| 4 |
+
"eval_f1": 0.9571847507331379,
|
| 5 |
+
"eval_loss": 0.04227477312088013,
|
| 6 |
+
"eval_precision": 0.953111963957951,
|
| 7 |
+
"eval_recall": 0.9612924941097274,
|
| 8 |
+
"eval_runtime": 6.2758,
|
| 9 |
+
"eval_samples": 3251,
|
| 10 |
+
"eval_samples_per_second": 518.019,
|
| 11 |
+
"eval_steps_per_second": 0.637
|
| 12 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"train_loss": 0.021953743980689482,
|
| 4 |
+
"train_runtime": 1347.1839,
|
| 5 |
+
"train_samples": 14042,
|
| 6 |
+
"train_samples_per_second": 104.232,
|
| 7 |
+
"train_steps_per_second": 1.633
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1105 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.9571847507331379,
|
| 3 |
+
"best_model_checkpoint": "twitter-roberta-base-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": 0.9385927339278065,
|
| 13 |
+
"eval_f1": 0.6587314799400699,
|
| 14 |
+
"eval_loss": 0.2062799483537674,
|
| 15 |
+
"eval_precision": 0.6516798418972332,
|
| 16 |
+
"eval_recall": 0.6659373948165601,
|
| 17 |
+
"eval_runtime": 5.9743,
|
| 18 |
+
"eval_samples_per_second": 544.169,
|
| 19 |
+
"eval_steps_per_second": 0.67,
|
| 20 |
+
"step": 25
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.23,
|
| 24 |
+
"eval_accuracy": 0.9771231649857871,
|
| 25 |
+
"eval_f1": 0.8565321055660609,
|
| 26 |
+
"eval_loss": 0.08096875995397568,
|
| 27 |
+
"eval_precision": 0.8373251888763864,
|
| 28 |
+
"eval_recall": 0.8766408616627398,
|
| 29 |
+
"eval_runtime": 6.1826,
|
| 30 |
+
"eval_samples_per_second": 525.833,
|
| 31 |
+
"eval_steps_per_second": 0.647,
|
| 32 |
+
"step": 50
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.34,
|
| 36 |
+
"eval_accuracy": 0.9826914839764807,
|
| 37 |
+
"eval_f1": 0.8996990972918756,
|
| 38 |
+
"eval_loss": 0.06511491537094116,
|
| 39 |
+
"eval_precision": 0.8937230156094321,
|
| 40 |
+
"eval_recall": 0.9057556378323797,
|
| 41 |
+
"eval_runtime": 6.0199,
|
| 42 |
+
"eval_samples_per_second": 540.041,
|
| 43 |
+
"eval_steps_per_second": 0.664,
|
| 44 |
+
"step": 75
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.45,
|
| 48 |
+
"eval_accuracy": 0.9849110237140298,
|
| 49 |
+
"eval_f1": 0.907388833166165,
|
| 50 |
+
"eval_loss": 0.05366199463605881,
|
| 51 |
+
"eval_precision": 0.9013616738625041,
|
| 52 |
+
"eval_recall": 0.9134971390104342,
|
| 53 |
+
"eval_runtime": 6.2998,
|
| 54 |
+
"eval_samples_per_second": 516.048,
|
| 55 |
+
"eval_steps_per_second": 0.635,
|
| 56 |
+
"step": 100
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.57,
|
| 60 |
+
"eval_accuracy": 0.986682761574705,
|
| 61 |
+
"eval_f1": 0.9170283806343906,
|
| 62 |
+
"eval_loss": 0.04637985676527023,
|
| 63 |
+
"eval_precision": 0.9097383239483272,
|
| 64 |
+
"eval_recall": 0.924436216762033,
|
| 65 |
+
"eval_runtime": 6.0114,
|
| 66 |
+
"eval_samples_per_second": 540.803,
|
| 67 |
+
"eval_steps_per_second": 0.665,
|
| 68 |
+
"step": 125
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.68,
|
| 72 |
+
"eval_accuracy": 0.9884544994353802,
|
| 73 |
+
"eval_f1": 0.9296410942859533,
|
| 74 |
+
"eval_loss": 0.042301442474126816,
|
| 75 |
+
"eval_precision": 0.9243054400266179,
|
| 76 |
+
"eval_recall": 0.9350387075058902,
|
| 77 |
+
"eval_runtime": 6.0918,
|
| 78 |
+
"eval_samples_per_second": 533.67,
|
| 79 |
+
"eval_steps_per_second": 0.657,
|
| 80 |
+
"step": 150
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.8,
|
| 84 |
+
"eval_accuracy": 0.9899731318873876,
|
| 85 |
+
"eval_f1": 0.934277384423157,
|
| 86 |
+
"eval_loss": 0.038059305399656296,
|
| 87 |
+
"eval_precision": 0.9249546429160481,
|
| 88 |
+
"eval_recall": 0.9437899697071693,
|
| 89 |
+
"eval_runtime": 6.2554,
|
| 90 |
+
"eval_samples_per_second": 519.713,
|
| 91 |
+
"eval_steps_per_second": 0.639,
|
| 92 |
+
"step": 175
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.91,
|
| 96 |
+
"eval_accuracy": 0.9895642693041549,
|
| 97 |
+
"eval_f1": 0.9354220481626531,
|
| 98 |
+
"eval_loss": 0.038840554654598236,
|
| 99 |
+
"eval_precision": 0.9263904934807724,
|
| 100 |
+
"eval_recall": 0.9446314372265231,
|
| 101 |
+
"eval_runtime": 6.0921,
|
| 102 |
+
"eval_samples_per_second": 533.639,
|
| 103 |
+
"eval_steps_per_second": 0.657,
|
| 104 |
+
"step": 200
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 1.02,
|
| 108 |
+
"eval_accuracy": 0.9898368443596434,
|
| 109 |
+
"eval_f1": 0.9384409501505521,
|
| 110 |
+
"eval_loss": 0.03943933546543121,
|
| 111 |
+
"eval_precision": 0.9328234120385767,
|
| 112 |
+
"eval_recall": 0.9441265567149109,
|
| 113 |
+
"eval_runtime": 6.2197,
|
| 114 |
+
"eval_samples_per_second": 522.692,
|
| 115 |
+
"eval_steps_per_second": 0.643,
|
| 116 |
+
"step": 225
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.14,
|
| 120 |
+
"eval_accuracy": 0.9895837389509754,
|
| 121 |
+
"eval_f1": 0.9402710389827672,
|
| 122 |
+
"eval_loss": 0.042259786278009415,
|
| 123 |
+
"eval_precision": 0.9347970725216235,
|
| 124 |
+
"eval_recall": 0.9458094917536183,
|
| 125 |
+
"eval_runtime": 6.2802,
|
| 126 |
+
"eval_samples_per_second": 517.656,
|
| 127 |
+
"eval_steps_per_second": 0.637,
|
| 128 |
+
"step": 250
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 1.25,
|
| 132 |
+
"eval_accuracy": 0.9892138156613839,
|
| 133 |
+
"eval_f1": 0.9354757720311323,
|
| 134 |
+
"eval_loss": 0.04323225095868111,
|
| 135 |
+
"eval_precision": 0.9304145163975363,
|
| 136 |
+
"eval_recall": 0.9405923931336251,
|
| 137 |
+
"eval_runtime": 6.1306,
|
| 138 |
+
"eval_samples_per_second": 530.294,
|
| 139 |
+
"eval_steps_per_second": 0.652,
|
| 140 |
+
"step": 275
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 1.36,
|
| 144 |
+
"eval_accuracy": 0.9900899497683112,
|
| 145 |
+
"eval_f1": 0.9432760787599497,
|
| 146 |
+
"eval_loss": 0.03824295476078987,
|
| 147 |
+
"eval_precision": 0.9392624728850325,
|
| 148 |
+
"eval_recall": 0.9473241332884551,
|
| 149 |
+
"eval_runtime": 6.3242,
|
| 150 |
+
"eval_samples_per_second": 514.058,
|
| 151 |
+
"eval_steps_per_second": 0.632,
|
| 152 |
+
"step": 300
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 1.48,
|
| 156 |
+
"eval_accuracy": 0.99005101047467,
|
| 157 |
+
"eval_f1": 0.9414020171709594,
|
| 158 |
+
"eval_loss": 0.0380987711250782,
|
| 159 |
+
"eval_precision": 0.932617671345995,
|
| 160 |
+
"eval_recall": 0.9503534163581285,
|
| 161 |
+
"eval_runtime": 6.1123,
|
| 162 |
+
"eval_samples_per_second": 531.88,
|
| 163 |
+
"eval_steps_per_second": 0.654,
|
| 164 |
+
"step": 325
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 1.59,
|
| 168 |
+
"eval_accuracy": 0.9902262372960554,
|
| 169 |
+
"eval_f1": 0.9429309339331834,
|
| 170 |
+
"eval_loss": 0.03868376836180687,
|
| 171 |
+
"eval_precision": 0.933674311169774,
|
| 172 |
+
"eval_recall": 0.9523729384045776,
|
| 173 |
+
"eval_runtime": 6.4317,
|
| 174 |
+
"eval_samples_per_second": 505.469,
|
| 175 |
+
"eval_steps_per_second": 0.622,
|
| 176 |
+
"step": 350
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 1.7,
|
| 180 |
+
"eval_accuracy": 0.990148358708773,
|
| 181 |
+
"eval_f1": 0.9439181825802665,
|
| 182 |
+
"eval_loss": 0.03645554929971695,
|
| 183 |
+
"eval_precision": 0.9403708034073827,
|
| 184 |
+
"eval_recall": 0.9474924267923258,
|
| 185 |
+
"eval_runtime": 6.1333,
|
| 186 |
+
"eval_samples_per_second": 530.054,
|
| 187 |
+
"eval_steps_per_second": 0.652,
|
| 188 |
+
"step": 375
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 1.82,
|
| 192 |
+
"eval_accuracy": 0.9905182819983646,
|
| 193 |
+
"eval_f1": 0.9473948735131512,
|
| 194 |
+
"eval_loss": 0.0382225401699543,
|
| 195 |
+
"eval_precision": 0.9431287525016678,
|
| 196 |
+
"eval_recall": 0.9516997643890945,
|
| 197 |
+
"eval_runtime": 6.1925,
|
| 198 |
+
"eval_samples_per_second": 524.987,
|
| 199 |
+
"eval_steps_per_second": 0.646,
|
| 200 |
+
"step": 400
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 1.93,
|
| 204 |
+
"eval_accuracy": 0.9903235855301585,
|
| 205 |
+
"eval_f1": 0.9460837582546184,
|
| 206 |
+
"eval_loss": 0.03733981028199196,
|
| 207 |
+
"eval_precision": 0.9398770968277694,
|
| 208 |
+
"eval_recall": 0.9523729384045776,
|
| 209 |
+
"eval_runtime": 6.4026,
|
| 210 |
+
"eval_samples_per_second": 507.766,
|
| 211 |
+
"eval_steps_per_second": 0.625,
|
| 212 |
+
"step": 425
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 2.05,
|
| 216 |
+
"eval_accuracy": 0.9910439624625209,
|
| 217 |
+
"eval_f1": 0.9497365559923059,
|
| 218 |
+
"eval_loss": 0.03666611388325691,
|
| 219 |
+
"eval_precision": 0.943973399833749,
|
| 220 |
+
"eval_recall": 0.9555705149781218,
|
| 221 |
+
"eval_runtime": 6.1803,
|
| 222 |
+
"eval_samples_per_second": 526.025,
|
| 223 |
+
"eval_steps_per_second": 0.647,
|
| 224 |
+
"step": 450
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 2.16,
|
| 228 |
+
"eval_accuracy": 0.99069350881975,
|
| 229 |
+
"eval_f1": 0.9474914433592119,
|
| 230 |
+
"eval_loss": 0.03960287570953369,
|
| 231 |
+
"eval_precision": 0.9400364419413616,
|
| 232 |
+
"eval_recall": 0.9550656344665096,
|
| 233 |
+
"eval_runtime": 6.4236,
|
| 234 |
+
"eval_samples_per_second": 506.1,
|
| 235 |
+
"eval_steps_per_second": 0.623,
|
| 236 |
+
"step": 475
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 2.27,
|
| 240 |
+
"learning_rate": 4.636363636363636e-05,
|
| 241 |
+
"loss": 0.0771,
|
| 242 |
+
"step": 500
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 2.27,
|
| 246 |
+
"eval_accuracy": 0.991238658930727,
|
| 247 |
+
"eval_f1": 0.9507813152836967,
|
| 248 |
+
"eval_loss": 0.035283163189888,
|
| 249 |
+
"eval_precision": 0.9442323651452282,
|
| 250 |
+
"eval_recall": 0.9574217435207001,
|
| 251 |
+
"eval_runtime": 6.2034,
|
| 252 |
+
"eval_samples_per_second": 524.07,
|
| 253 |
+
"eval_steps_per_second": 0.645,
|
| 254 |
+
"step": 500
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 2.39,
|
| 258 |
+
"eval_accuracy": 0.9905572212920057,
|
| 259 |
+
"eval_f1": 0.9453602209020165,
|
| 260 |
+
"eval_loss": 0.03937030956149101,
|
| 261 |
+
"eval_precision": 0.9400898652021967,
|
| 262 |
+
"eval_recall": 0.9506900033658701,
|
| 263 |
+
"eval_runtime": 6.0362,
|
| 264 |
+
"eval_samples_per_second": 538.581,
|
| 265 |
+
"eval_steps_per_second": 0.663,
|
| 266 |
+
"step": 525
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 2.5,
|
| 270 |
+
"eval_accuracy": 0.9909660838752385,
|
| 271 |
+
"eval_f1": 0.9484536082474228,
|
| 272 |
+
"eval_loss": 0.036998968571424484,
|
| 273 |
+
"eval_precision": 0.9447320086825848,
|
| 274 |
+
"eval_recall": 0.9522046449007069,
|
| 275 |
+
"eval_runtime": 6.4336,
|
| 276 |
+
"eval_samples_per_second": 505.317,
|
| 277 |
+
"eval_steps_per_second": 0.622,
|
| 278 |
+
"step": 550
|
| 279 |
+
},
|
| 280 |
+
{
|
| 281 |
+
"epoch": 2.61,
|
| 282 |
+
"eval_accuracy": 0.9908492659943149,
|
| 283 |
+
"eval_f1": 0.9472013366750209,
|
| 284 |
+
"eval_loss": 0.035233963280916214,
|
| 285 |
+
"eval_precision": 0.940444591904446,
|
| 286 |
+
"eval_recall": 0.9540558734432851,
|
| 287 |
+
"eval_runtime": 6.3271,
|
| 288 |
+
"eval_samples_per_second": 513.819,
|
| 289 |
+
"eval_steps_per_second": 0.632,
|
| 290 |
+
"step": 575
|
| 291 |
+
},
|
| 292 |
+
{
|
| 293 |
+
"epoch": 2.73,
|
| 294 |
+
"eval_accuracy": 0.9908103267006737,
|
| 295 |
+
"eval_f1": 0.9448281601065159,
|
| 296 |
+
"eval_loss": 0.0385683998465538,
|
| 297 |
+
"eval_precision": 0.9344855967078189,
|
| 298 |
+
"eval_recall": 0.9554022214742511,
|
| 299 |
+
"eval_runtime": 6.4264,
|
| 300 |
+
"eval_samples_per_second": 505.878,
|
| 301 |
+
"eval_steps_per_second": 0.622,
|
| 302 |
+
"step": 600
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 2.84,
|
| 306 |
+
"eval_accuracy": 0.9916280518671391,
|
| 307 |
+
"eval_f1": 0.9501544627202138,
|
| 308 |
+
"eval_loss": 0.036591168493032455,
|
| 309 |
+
"eval_precision": 0.9428334714167357,
|
| 310 |
+
"eval_recall": 0.9575900370245709,
|
| 311 |
+
"eval_runtime": 6.0976,
|
| 312 |
+
"eval_samples_per_second": 533.163,
|
| 313 |
+
"eval_steps_per_second": 0.656,
|
| 314 |
+
"step": 625
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 2.95,
|
| 318 |
+
"eval_accuracy": 0.9913360071648301,
|
| 319 |
+
"eval_f1": 0.948574295509658,
|
| 320 |
+
"eval_loss": 0.035347189754247665,
|
| 321 |
+
"eval_precision": 0.9426624563736081,
|
| 322 |
+
"eval_recall": 0.9545607539548974,
|
| 323 |
+
"eval_runtime": 6.1569,
|
| 324 |
+
"eval_samples_per_second": 528.025,
|
| 325 |
+
"eval_steps_per_second": 0.65,
|
| 326 |
+
"step": 650
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 3.07,
|
| 330 |
+
"eval_accuracy": 0.9911023714029827,
|
| 331 |
+
"eval_f1": 0.9477730425336343,
|
| 332 |
+
"eval_loss": 0.035899706184864044,
|
| 333 |
+
"eval_precision": 0.9412448132780084,
|
| 334 |
+
"eval_recall": 0.9543924604510265,
|
| 335 |
+
"eval_runtime": 6.3818,
|
| 336 |
+
"eval_samples_per_second": 509.418,
|
| 337 |
+
"eval_steps_per_second": 0.627,
|
| 338 |
+
"step": 675
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 3.18,
|
| 342 |
+
"eval_accuracy": 0.9919979751567306,
|
| 343 |
+
"eval_f1": 0.9534164087981936,
|
| 344 |
+
"eval_loss": 0.0355631485581398,
|
| 345 |
+
"eval_precision": 0.9476309226932669,
|
| 346 |
+
"eval_recall": 0.9592729720632783,
|
| 347 |
+
"eval_runtime": 6.1016,
|
| 348 |
+
"eval_samples_per_second": 532.815,
|
| 349 |
+
"eval_steps_per_second": 0.656,
|
| 350 |
+
"step": 700
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 3.3,
|
| 354 |
+
"eval_accuracy": 0.9918422179821658,
|
| 355 |
+
"eval_f1": 0.953465015065283,
|
| 356 |
+
"eval_loss": 0.03445196524262428,
|
| 357 |
+
"eval_precision": 0.9483849483849484,
|
| 358 |
+
"eval_recall": 0.9585997980477954,
|
| 359 |
+
"eval_runtime": 6.4133,
|
| 360 |
+
"eval_samples_per_second": 506.912,
|
| 361 |
+
"eval_steps_per_second": 0.624,
|
| 362 |
+
"step": 725
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 3.41,
|
| 366 |
+
"eval_accuracy": 0.9915696429266773,
|
| 367 |
+
"eval_f1": 0.9491893698813304,
|
| 368 |
+
"eval_loss": 0.03445504605770111,
|
| 369 |
+
"eval_precision": 0.9427290836653387,
|
| 370 |
+
"eval_recall": 0.9557388084819926,
|
| 371 |
+
"eval_runtime": 6.2039,
|
| 372 |
+
"eval_samples_per_second": 524.023,
|
| 373 |
+
"eval_steps_per_second": 0.645,
|
| 374 |
+
"step": 750
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 3.52,
|
| 378 |
+
"eval_accuracy": 0.9913944161052919,
|
| 379 |
+
"eval_f1": 0.9478246374395733,
|
| 380 |
+
"eval_loss": 0.03635535016655922,
|
| 381 |
+
"eval_precision": 0.9389035667107001,
|
| 382 |
+
"eval_recall": 0.9569168630090878,
|
| 383 |
+
"eval_runtime": 6.2115,
|
| 384 |
+
"eval_samples_per_second": 523.384,
|
| 385 |
+
"eval_steps_per_second": 0.644,
|
| 386 |
+
"step": 775
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 3.64,
|
| 390 |
+
"eval_accuracy": 0.9914917643393949,
|
| 391 |
+
"eval_f1": 0.95067189717052,
|
| 392 |
+
"eval_loss": 0.03599384054541588,
|
| 393 |
+
"eval_precision": 0.943036926643484,
|
| 394 |
+
"eval_recall": 0.9584315045439246,
|
| 395 |
+
"eval_runtime": 6.3798,
|
| 396 |
+
"eval_samples_per_second": 509.575,
|
| 397 |
+
"eval_steps_per_second": 0.627,
|
| 398 |
+
"step": 800
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 3.75,
|
| 402 |
+
"eval_accuracy": 0.9914528250457537,
|
| 403 |
+
"eval_f1": 0.950514945993469,
|
| 404 |
+
"eval_loss": 0.03866244852542877,
|
| 405 |
+
"eval_precision": 0.9458423596067322,
|
| 406 |
+
"eval_recall": 0.9552339279703803,
|
| 407 |
+
"eval_runtime": 6.1387,
|
| 408 |
+
"eval_samples_per_second": 529.592,
|
| 409 |
+
"eval_steps_per_second": 0.652,
|
| 410 |
+
"step": 825
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 3.86,
|
| 414 |
+
"eval_accuracy": 0.9917059304544216,
|
| 415 |
+
"eval_f1": 0.9521419009370816,
|
| 416 |
+
"eval_loss": 0.034695982933044434,
|
| 417 |
+
"eval_precision": 0.9467554076539102,
|
| 418 |
+
"eval_recall": 0.9575900370245709,
|
| 419 |
+
"eval_runtime": 6.3766,
|
| 420 |
+
"eval_samples_per_second": 509.831,
|
| 421 |
+
"eval_steps_per_second": 0.627,
|
| 422 |
+
"step": 850
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 3.98,
|
| 426 |
+
"eval_accuracy": 0.9915307036330361,
|
| 427 |
+
"eval_f1": 0.9509402423735897,
|
| 428 |
+
"eval_loss": 0.03567422926425934,
|
| 429 |
+
"eval_precision": 0.9445459073551387,
|
| 430 |
+
"eval_recall": 0.9574217435207001,
|
| 431 |
+
"eval_runtime": 6.1035,
|
| 432 |
+
"eval_samples_per_second": 532.648,
|
| 433 |
+
"eval_steps_per_second": 0.655,
|
| 434 |
+
"step": 875
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"epoch": 4.09,
|
| 438 |
+
"eval_accuracy": 0.9918032786885246,
|
| 439 |
+
"eval_f1": 0.9520702634880803,
|
| 440 |
+
"eval_loss": 0.03815401718020439,
|
| 441 |
+
"eval_precision": 0.9464493597206054,
|
| 442 |
+
"eval_recall": 0.9577583305284416,
|
| 443 |
+
"eval_runtime": 6.1453,
|
| 444 |
+
"eval_samples_per_second": 529.025,
|
| 445 |
+
"eval_steps_per_second": 0.651,
|
| 446 |
+
"step": 900
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"epoch": 4.2,
|
| 450 |
+
"eval_accuracy": 0.9917643393948834,
|
| 451 |
+
"eval_f1": 0.9518385124382277,
|
| 452 |
+
"eval_loss": 0.03912338241934776,
|
| 453 |
+
"eval_precision": 0.9474737368684342,
|
| 454 |
+
"eval_recall": 0.9562436889936049,
|
| 455 |
+
"eval_runtime": 6.3668,
|
| 456 |
+
"eval_samples_per_second": 510.616,
|
| 457 |
+
"eval_steps_per_second": 0.628,
|
| 458 |
+
"step": 925
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 4.32,
|
| 462 |
+
"eval_accuracy": 0.9911802499902652,
|
| 463 |
+
"eval_f1": 0.9506493506493505,
|
| 464 |
+
"eval_loss": 0.0428401380777359,
|
| 465 |
+
"eval_precision": 0.9466043717670616,
|
| 466 |
+
"eval_recall": 0.9547290474587681,
|
| 467 |
+
"eval_runtime": 6.2553,
|
| 468 |
+
"eval_samples_per_second": 519.717,
|
| 469 |
+
"eval_steps_per_second": 0.639,
|
| 470 |
+
"step": 950
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 4.43,
|
| 474 |
+
"eval_accuracy": 0.9913360071648301,
|
| 475 |
+
"eval_f1": 0.9506028131279304,
|
| 476 |
+
"eval_loss": 0.04035865515470505,
|
| 477 |
+
"eval_precision": 0.9458513828723759,
|
| 478 |
+
"eval_recall": 0.9554022214742511,
|
| 479 |
+
"eval_runtime": 6.4802,
|
| 480 |
+
"eval_samples_per_second": 501.682,
|
| 481 |
+
"eval_steps_per_second": 0.617,
|
| 482 |
+
"step": 975
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 4.55,
|
| 486 |
+
"learning_rate": 3.2727272727272725e-05,
|
| 487 |
+
"loss": 0.0118,
|
| 488 |
+
"step": 1000
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 4.55,
|
| 492 |
+
"eval_accuracy": 0.9908687356411354,
|
| 493 |
+
"eval_f1": 0.9461397365349341,
|
| 494 |
+
"eval_loss": 0.04028761386871338,
|
| 495 |
+
"eval_precision": 0.9375413086582948,
|
| 496 |
+
"eval_recall": 0.9548973409626389,
|
| 497 |
+
"eval_runtime": 6.2199,
|
| 498 |
+
"eval_samples_per_second": 522.675,
|
| 499 |
+
"eval_steps_per_second": 0.643,
|
| 500 |
+
"step": 1000
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 4.66,
|
| 504 |
+
"eval_accuracy": 0.9919200965694482,
|
| 505 |
+
"eval_f1": 0.9533852205205458,
|
| 506 |
+
"eval_loss": 0.036938250064849854,
|
| 507 |
+
"eval_precision": 0.9482270684201765,
|
| 508 |
+
"eval_recall": 0.9585997980477954,
|
| 509 |
+
"eval_runtime": 6.0752,
|
| 510 |
+
"eval_samples_per_second": 535.129,
|
| 511 |
+
"eval_steps_per_second": 0.658,
|
| 512 |
+
"step": 1025
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 4.77,
|
| 516 |
+
"eval_accuracy": 0.9918032786885246,
|
| 517 |
+
"eval_f1": 0.9520227348712804,
|
| 518 |
+
"eval_loss": 0.037420712411403656,
|
| 519 |
+
"eval_precision": 0.9456991032879442,
|
| 520 |
+
"eval_recall": 0.9584315045439246,
|
| 521 |
+
"eval_runtime": 6.266,
|
| 522 |
+
"eval_samples_per_second": 518.831,
|
| 523 |
+
"eval_steps_per_second": 0.638,
|
| 524 |
+
"step": 1050
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 4.89,
|
| 528 |
+
"eval_accuracy": 0.9922900198590398,
|
| 529 |
+
"eval_f1": 0.9538745387453874,
|
| 530 |
+
"eval_loss": 0.035940222442150116,
|
| 531 |
+
"eval_precision": 0.9506853895018389,
|
| 532 |
+
"eval_recall": 0.9570851565129586,
|
| 533 |
+
"eval_runtime": 6.1032,
|
| 534 |
+
"eval_samples_per_second": 532.673,
|
| 535 |
+
"eval_steps_per_second": 0.655,
|
| 536 |
+
"step": 1075
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 5.0,
|
| 540 |
+
"eval_accuracy": 0.9919006269226276,
|
| 541 |
+
"eval_f1": 0.9523093627328155,
|
| 542 |
+
"eval_loss": 0.03730743005871773,
|
| 543 |
+
"eval_precision": 0.9452827060189023,
|
| 544 |
+
"eval_recall": 0.9594412655671491,
|
| 545 |
+
"eval_runtime": 6.3652,
|
| 546 |
+
"eval_samples_per_second": 510.745,
|
| 547 |
+
"eval_steps_per_second": 0.628,
|
| 548 |
+
"step": 1100
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 5.11,
|
| 552 |
+
"eval_accuracy": 0.9923873680931428,
|
| 553 |
+
"eval_f1": 0.9546215673141327,
|
| 554 |
+
"eval_loss": 0.037039875984191895,
|
| 555 |
+
"eval_precision": 0.9498500499833389,
|
| 556 |
+
"eval_recall": 0.9594412655671491,
|
| 557 |
+
"eval_runtime": 6.169,
|
| 558 |
+
"eval_samples_per_second": 526.993,
|
| 559 |
+
"eval_steps_per_second": 0.648,
|
| 560 |
+
"step": 1125
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 5.23,
|
| 564 |
+
"eval_accuracy": 0.9922316109185779,
|
| 565 |
+
"eval_f1": 0.9555313625324512,
|
| 566 |
+
"eval_loss": 0.038839343935251236,
|
| 567 |
+
"eval_precision": 0.9509918319719953,
|
| 568 |
+
"eval_recall": 0.9601144395826321,
|
| 569 |
+
"eval_runtime": 6.0785,
|
| 570 |
+
"eval_samples_per_second": 534.833,
|
| 571 |
+
"eval_steps_per_second": 0.658,
|
| 572 |
+
"step": 1150
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 5.34,
|
| 576 |
+
"eval_accuracy": 0.9919785055099101,
|
| 577 |
+
"eval_f1": 0.9522212908633696,
|
| 578 |
+
"eval_loss": 0.03950090706348419,
|
| 579 |
+
"eval_precision": 0.948563794255177,
|
| 580 |
+
"eval_recall": 0.9559071019858634,
|
| 581 |
+
"eval_runtime": 6.4405,
|
| 582 |
+
"eval_samples_per_second": 504.775,
|
| 583 |
+
"eval_steps_per_second": 0.621,
|
| 584 |
+
"step": 1175
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 5.45,
|
| 588 |
+
"eval_accuracy": 0.9923678984463222,
|
| 589 |
+
"eval_f1": 0.9542866711319491,
|
| 590 |
+
"eval_loss": 0.03910110890865326,
|
| 591 |
+
"eval_precision": 0.949516827724092,
|
| 592 |
+
"eval_recall": 0.9591046785594076,
|
| 593 |
+
"eval_runtime": 6.1207,
|
| 594 |
+
"eval_samples_per_second": 531.15,
|
| 595 |
+
"eval_steps_per_second": 0.654,
|
| 596 |
+
"step": 1200
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 5.57,
|
| 600 |
+
"eval_accuracy": 0.9923289591526809,
|
| 601 |
+
"eval_f1": 0.9552313883299799,
|
| 602 |
+
"eval_loss": 0.037824589759111404,
|
| 603 |
+
"eval_precision": 0.9517206815903776,
|
| 604 |
+
"eval_recall": 0.9587680915516661,
|
| 605 |
+
"eval_runtime": 6.1605,
|
| 606 |
+
"eval_samples_per_second": 527.72,
|
| 607 |
+
"eval_steps_per_second": 0.649,
|
| 608 |
+
"step": 1225
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 5.68,
|
| 612 |
+
"eval_accuracy": 0.9923678984463222,
|
| 613 |
+
"eval_f1": 0.956470785200067,
|
| 614 |
+
"eval_loss": 0.038752481341362,
|
| 615 |
+
"eval_precision": 0.9515323117921386,
|
| 616 |
+
"eval_recall": 0.9614607876135981,
|
| 617 |
+
"eval_runtime": 6.4555,
|
| 618 |
+
"eval_samples_per_second": 503.603,
|
| 619 |
+
"eval_steps_per_second": 0.62,
|
| 620 |
+
"step": 1250
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 5.8,
|
| 624 |
+
"eval_accuracy": 0.9924457770336046,
|
| 625 |
+
"eval_f1": 0.9560485558811219,
|
| 626 |
+
"eval_loss": 0.03838498517870903,
|
| 627 |
+
"eval_precision": 0.9511910711311011,
|
| 628 |
+
"eval_recall": 0.9609559071019859,
|
| 629 |
+
"eval_runtime": 6.1584,
|
| 630 |
+
"eval_samples_per_second": 527.9,
|
| 631 |
+
"eval_steps_per_second": 0.65,
|
| 632 |
+
"step": 1275
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 5.91,
|
| 636 |
+
"eval_accuracy": 0.9924068377399634,
|
| 637 |
+
"eval_f1": 0.9571045576407506,
|
| 638 |
+
"eval_loss": 0.03946718946099281,
|
| 639 |
+
"eval_precision": 0.9529529529529529,
|
| 640 |
+
"eval_recall": 0.9612924941097274,
|
| 641 |
+
"eval_runtime": 6.2271,
|
| 642 |
+
"eval_samples_per_second": 522.069,
|
| 643 |
+
"eval_steps_per_second": 0.642,
|
| 644 |
+
"step": 1300
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 6.02,
|
| 648 |
+
"eval_accuracy": 0.9919200965694482,
|
| 649 |
+
"eval_f1": 0.9533869885982562,
|
| 650 |
+
"eval_loss": 0.04075756296515465,
|
| 651 |
+
"eval_precision": 0.9498830604744404,
|
| 652 |
+
"eval_recall": 0.9569168630090878,
|
| 653 |
+
"eval_runtime": 6.3886,
|
| 654 |
+
"eval_samples_per_second": 508.877,
|
| 655 |
+
"eval_steps_per_second": 0.626,
|
| 656 |
+
"step": 1325
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 6.14,
|
| 660 |
+
"eval_accuracy": 0.9921732019781161,
|
| 661 |
+
"eval_f1": 0.9547998997409975,
|
| 662 |
+
"eval_loss": 0.04120802879333496,
|
| 663 |
+
"eval_precision": 0.948067031690725,
|
| 664 |
+
"eval_recall": 0.9616290811174689,
|
| 665 |
+
"eval_runtime": 6.1801,
|
| 666 |
+
"eval_samples_per_second": 526.047,
|
| 667 |
+
"eval_steps_per_second": 0.647,
|
| 668 |
+
"step": 1350
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 6.25,
|
| 672 |
+
"eval_accuracy": 0.9923873680931428,
|
| 673 |
+
"eval_f1": 0.9555667337357477,
|
| 674 |
+
"eval_loss": 0.0413355678319931,
|
| 675 |
+
"eval_precision": 0.952054794520548,
|
| 676 |
+
"eval_recall": 0.9591046785594076,
|
| 677 |
+
"eval_runtime": 6.3646,
|
| 678 |
+
"eval_samples_per_second": 510.793,
|
| 679 |
+
"eval_steps_per_second": 0.628,
|
| 680 |
+
"step": 1375
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 6.36,
|
| 684 |
+
"eval_accuracy": 0.9917448697480628,
|
| 685 |
+
"eval_f1": 0.9525004181301221,
|
| 686 |
+
"eval_loss": 0.04124680161476135,
|
| 687 |
+
"eval_precision": 0.9466422872340425,
|
| 688 |
+
"eval_recall": 0.9584315045439246,
|
| 689 |
+
"eval_runtime": 6.1566,
|
| 690 |
+
"eval_samples_per_second": 528.052,
|
| 691 |
+
"eval_steps_per_second": 0.65,
|
| 692 |
+
"step": 1400
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 6.48,
|
| 696 |
+
"eval_accuracy": 0.9920953233908337,
|
| 697 |
+
"eval_f1": 0.9555611348229978,
|
| 698 |
+
"eval_loss": 0.040540654212236404,
|
| 699 |
+
"eval_precision": 0.9503912102547029,
|
| 700 |
+
"eval_recall": 0.9607876135981152,
|
| 701 |
+
"eval_runtime": 6.4312,
|
| 702 |
+
"eval_samples_per_second": 505.501,
|
| 703 |
+
"eval_steps_per_second": 0.622,
|
| 704 |
+
"step": 1425
|
| 705 |
+
},
|
| 706 |
+
{
|
| 707 |
+
"epoch": 6.59,
|
| 708 |
+
"eval_accuracy": 0.9925041859740664,
|
| 709 |
+
"eval_f1": 0.9565508580996233,
|
| 710 |
+
"eval_loss": 0.04004018381237984,
|
| 711 |
+
"eval_precision": 0.9516908212560387,
|
| 712 |
+
"eval_recall": 0.9614607876135981,
|
| 713 |
+
"eval_runtime": 6.14,
|
| 714 |
+
"eval_samples_per_second": 529.477,
|
| 715 |
+
"eval_steps_per_second": 0.651,
|
| 716 |
+
"step": 1450
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 6.7,
|
| 720 |
+
"eval_accuracy": 0.9922900198590398,
|
| 721 |
+
"eval_f1": 0.9551813688531458,
|
| 722 |
+
"eval_loss": 0.039815619587898254,
|
| 723 |
+
"eval_precision": 0.9509591326105088,
|
| 724 |
+
"eval_recall": 0.9594412655671491,
|
| 725 |
+
"eval_runtime": 6.1824,
|
| 726 |
+
"eval_samples_per_second": 525.848,
|
| 727 |
+
"eval_steps_per_second": 0.647,
|
| 728 |
+
"step": 1475
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 6.82,
|
| 732 |
+
"learning_rate": 1.909090909090909e-05,
|
| 733 |
+
"loss": 0.0049,
|
| 734 |
+
"step": 1500
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 6.82,
|
| 738 |
+
"eval_accuracy": 0.9925431252677076,
|
| 739 |
+
"eval_f1": 0.956871283812076,
|
| 740 |
+
"eval_loss": 0.03951007500290871,
|
| 741 |
+
"eval_precision": 0.9523253875645941,
|
| 742 |
+
"eval_recall": 0.9614607876135981,
|
| 743 |
+
"eval_runtime": 6.3696,
|
| 744 |
+
"eval_samples_per_second": 510.391,
|
| 745 |
+
"eval_steps_per_second": 0.628,
|
| 746 |
+
"step": 1500
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"epoch": 6.93,
|
| 750 |
+
"eval_accuracy": 0.9927183520890931,
|
| 751 |
+
"eval_f1": 0.9571476397723468,
|
| 752 |
+
"eval_loss": 0.03918459266424179,
|
| 753 |
+
"eval_precision": 0.952047952047952,
|
| 754 |
+
"eval_recall": 0.9623022551329519,
|
| 755 |
+
"eval_runtime": 6.0615,
|
| 756 |
+
"eval_samples_per_second": 536.339,
|
| 757 |
+
"eval_steps_per_second": 0.66,
|
| 758 |
+
"step": 1525
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 7.05,
|
| 762 |
+
"eval_accuracy": 0.9923484287995016,
|
| 763 |
+
"eval_f1": 0.9551738583996647,
|
| 764 |
+
"eval_loss": 0.03897989168763161,
|
| 765 |
+
"eval_precision": 0.9511096278992157,
|
| 766 |
+
"eval_recall": 0.9592729720632783,
|
| 767 |
+
"eval_runtime": 6.0213,
|
| 768 |
+
"eval_samples_per_second": 539.918,
|
| 769 |
+
"eval_steps_per_second": 0.664,
|
| 770 |
+
"step": 1550
|
| 771 |
+
},
|
| 772 |
+
{
|
| 773 |
+
"epoch": 7.16,
|
| 774 |
+
"eval_accuracy": 0.9925431252677076,
|
| 775 |
+
"eval_f1": 0.9565363034921699,
|
| 776 |
+
"eval_loss": 0.03931865468621254,
|
| 777 |
+
"eval_precision": 0.9519919986664445,
|
| 778 |
+
"eval_recall": 0.9611242006058566,
|
| 779 |
+
"eval_runtime": 6.2836,
|
| 780 |
+
"eval_samples_per_second": 517.381,
|
| 781 |
+
"eval_steps_per_second": 0.637,
|
| 782 |
+
"step": 1575
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 7.27,
|
| 786 |
+
"eval_accuracy": 0.9925236556208871,
|
| 787 |
+
"eval_f1": 0.9562233196618398,
|
| 788 |
+
"eval_loss": 0.03892602398991585,
|
| 789 |
+
"eval_precision": 0.9512073272273106,
|
| 790 |
+
"eval_recall": 0.9612924941097274,
|
| 791 |
+
"eval_runtime": 6.0872,
|
| 792 |
+
"eval_samples_per_second": 534.07,
|
| 793 |
+
"eval_steps_per_second": 0.657,
|
| 794 |
+
"step": 1600
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 7.39,
|
| 798 |
+
"eval_accuracy": 0.9923678984463222,
|
| 799 |
+
"eval_f1": 0.9565435820145692,
|
| 800 |
+
"eval_loss": 0.0404973067343235,
|
| 801 |
+
"eval_precision": 0.9518413597733711,
|
| 802 |
+
"eval_recall": 0.9612924941097274,
|
| 803 |
+
"eval_runtime": 6.4196,
|
| 804 |
+
"eval_samples_per_second": 506.419,
|
| 805 |
+
"eval_steps_per_second": 0.623,
|
| 806 |
+
"step": 1625
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 7.5,
|
| 810 |
+
"eval_accuracy": 0.9924847163272458,
|
| 811 |
+
"eval_f1": 0.9558737335677803,
|
| 812 |
+
"eval_loss": 0.04104885458946228,
|
| 813 |
+
"eval_precision": 0.9511748041993001,
|
| 814 |
+
"eval_recall": 0.9606193200942443,
|
| 815 |
+
"eval_runtime": 6.0828,
|
| 816 |
+
"eval_samples_per_second": 534.456,
|
| 817 |
+
"eval_steps_per_second": 0.658,
|
| 818 |
+
"step": 1650
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 7.61,
|
| 822 |
+
"eval_accuracy": 0.9925431252677076,
|
| 823 |
+
"eval_f1": 0.9569442117607639,
|
| 824 |
+
"eval_loss": 0.04079214483499527,
|
| 825 |
+
"eval_precision": 0.95263509006004,
|
| 826 |
+
"eval_recall": 0.9612924941097274,
|
| 827 |
+
"eval_runtime": 6.2616,
|
| 828 |
+
"eval_samples_per_second": 519.196,
|
| 829 |
+
"eval_steps_per_second": 0.639,
|
| 830 |
+
"step": 1675
|
| 831 |
+
},
|
| 832 |
+
{
|
| 833 |
+
"epoch": 7.73,
|
| 834 |
+
"eval_accuracy": 0.9921732019781161,
|
| 835 |
+
"eval_f1": 0.9545302574389837,
|
| 836 |
+
"eval_loss": 0.043598320335149765,
|
| 837 |
+
"eval_precision": 0.9481899701095982,
|
| 838 |
+
"eval_recall": 0.9609559071019859,
|
| 839 |
+
"eval_runtime": 6.3776,
|
| 840 |
+
"eval_samples_per_second": 509.751,
|
| 841 |
+
"eval_steps_per_second": 0.627,
|
| 842 |
+
"step": 1700
|
| 843 |
+
},
|
| 844 |
+
{
|
| 845 |
+
"epoch": 7.84,
|
| 846 |
+
"eval_accuracy": 0.9924068377399634,
|
| 847 |
+
"eval_f1": 0.9559548683660676,
|
| 848 |
+
"eval_loss": 0.0419035442173481,
|
| 849 |
+
"eval_precision": 0.9495268138801262,
|
| 850 |
+
"eval_recall": 0.9624705486368226,
|
| 851 |
+
"eval_runtime": 6.107,
|
| 852 |
+
"eval_samples_per_second": 532.344,
|
| 853 |
+
"eval_steps_per_second": 0.655,
|
| 854 |
+
"step": 1725
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 7.95,
|
| 858 |
+
"eval_accuracy": 0.9925625949145283,
|
| 859 |
+
"eval_f1": 0.957126109529392,
|
| 860 |
+
"eval_loss": 0.04293340817093849,
|
| 861 |
+
"eval_precision": 0.9525,
|
| 862 |
+
"eval_recall": 0.9617973746213396,
|
| 863 |
+
"eval_runtime": 6.1631,
|
| 864 |
+
"eval_samples_per_second": 527.491,
|
| 865 |
+
"eval_steps_per_second": 0.649,
|
| 866 |
+
"step": 1750
|
| 867 |
+
},
|
| 868 |
+
{
|
| 869 |
+
"epoch": 8.07,
|
| 870 |
+
"eval_accuracy": 0.9924068377399634,
|
| 871 |
+
"eval_f1": 0.9561506276150629,
|
| 872 |
+
"eval_loss": 0.041886329650878906,
|
| 873 |
+
"eval_precision": 0.9508988015978695,
|
| 874 |
+
"eval_recall": 0.9614607876135981,
|
| 875 |
+
"eval_runtime": 6.3831,
|
| 876 |
+
"eval_samples_per_second": 509.311,
|
| 877 |
+
"eval_steps_per_second": 0.627,
|
| 878 |
+
"step": 1775
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 8.18,
|
| 882 |
+
"eval_accuracy": 0.9922900198590398,
|
| 883 |
+
"eval_f1": 0.9555313625324512,
|
| 884 |
+
"eval_loss": 0.04220963269472122,
|
| 885 |
+
"eval_precision": 0.9509918319719953,
|
| 886 |
+
"eval_recall": 0.9601144395826321,
|
| 887 |
+
"eval_runtime": 6.1467,
|
| 888 |
+
"eval_samples_per_second": 528.904,
|
| 889 |
+
"eval_steps_per_second": 0.651,
|
| 890 |
+
"step": 1800
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 8.3,
|
| 894 |
+
"eval_accuracy": 0.9923873680931428,
|
| 895 |
+
"eval_f1": 0.9561793045664014,
|
| 896 |
+
"eval_loss": 0.041702426970005035,
|
| 897 |
+
"eval_precision": 0.9521107959285834,
|
| 898 |
+
"eval_recall": 0.9602827330865029,
|
| 899 |
+
"eval_runtime": 6.4978,
|
| 900 |
+
"eval_samples_per_second": 500.322,
|
| 901 |
+
"eval_steps_per_second": 0.616,
|
| 902 |
+
"step": 1825
|
| 903 |
+
},
|
| 904 |
+
{
|
| 905 |
+
"epoch": 8.41,
|
| 906 |
+
"eval_accuracy": 0.9924652466804252,
|
| 907 |
+
"eval_f1": 0.9570171763720151,
|
| 908 |
+
"eval_loss": 0.04150727763772011,
|
| 909 |
+
"eval_precision": 0.952945102619723,
|
| 910 |
+
"eval_recall": 0.9611242006058566,
|
| 911 |
+
"eval_runtime": 6.1888,
|
| 912 |
+
"eval_samples_per_second": 525.307,
|
| 913 |
+
"eval_steps_per_second": 0.646,
|
| 914 |
+
"step": 1850
|
| 915 |
+
},
|
| 916 |
+
{
|
| 917 |
+
"epoch": 8.52,
|
| 918 |
+
"eval_accuracy": 0.9924263073867841,
|
| 919 |
+
"eval_f1": 0.9566965407488064,
|
| 920 |
+
"eval_loss": 0.04162871465086937,
|
| 921 |
+
"eval_precision": 0.952309488077372,
|
| 922 |
+
"eval_recall": 0.9611242006058566,
|
| 923 |
+
"eval_runtime": 6.155,
|
| 924 |
+
"eval_samples_per_second": 528.192,
|
| 925 |
+
"eval_steps_per_second": 0.65,
|
| 926 |
+
"step": 1875
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 8.64,
|
| 930 |
+
"eval_accuracy": 0.9922316109185779,
|
| 931 |
+
"eval_f1": 0.9555611348229978,
|
| 932 |
+
"eval_loss": 0.04189736768603325,
|
| 933 |
+
"eval_precision": 0.9503912102547029,
|
| 934 |
+
"eval_recall": 0.9607876135981152,
|
| 935 |
+
"eval_runtime": 6.4706,
|
| 936 |
+
"eval_samples_per_second": 502.422,
|
| 937 |
+
"eval_steps_per_second": 0.618,
|
| 938 |
+
"step": 1900
|
| 939 |
+
},
|
| 940 |
+
{
|
| 941 |
+
"epoch": 8.75,
|
| 942 |
+
"eval_accuracy": 0.9923873680931428,
|
| 943 |
+
"eval_f1": 0.9564489112227805,
|
| 944 |
+
"eval_loss": 0.04165585711598396,
|
| 945 |
+
"eval_precision": 0.9519839946648883,
|
| 946 |
+
"eval_recall": 0.9609559071019859,
|
| 947 |
+
"eval_runtime": 6.1453,
|
| 948 |
+
"eval_samples_per_second": 529.021,
|
| 949 |
+
"eval_steps_per_second": 0.651,
|
| 950 |
+
"step": 1925
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 8.86,
|
| 954 |
+
"eval_accuracy": 0.9926015342081694,
|
| 955 |
+
"eval_f1": 0.9577818730105546,
|
| 956 |
+
"eval_loss": 0.041886311024427414,
|
| 957 |
+
"eval_precision": 0.9534689793195463,
|
| 958 |
+
"eval_recall": 0.9621339616290812,
|
| 959 |
+
"eval_runtime": 6.3199,
|
| 960 |
+
"eval_samples_per_second": 514.406,
|
| 961 |
+
"eval_steps_per_second": 0.633,
|
| 962 |
+
"step": 1950
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 8.98,
|
| 966 |
+
"eval_accuracy": 0.9926599431486313,
|
| 967 |
+
"eval_f1": 0.9575341318368372,
|
| 968 |
+
"eval_loss": 0.04215257614850998,
|
| 969 |
+
"eval_precision": 0.9531432382858096,
|
| 970 |
+
"eval_recall": 0.9619656681252103,
|
| 971 |
+
"eval_runtime": 6.4751,
|
| 972 |
+
"eval_samples_per_second": 502.081,
|
| 973 |
+
"eval_steps_per_second": 0.618,
|
| 974 |
+
"step": 1975
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 9.09,
|
| 978 |
+
"learning_rate": 5.4545454545454545e-06,
|
| 979 |
+
"loss": 0.0022,
|
| 980 |
+
"step": 2000
|
| 981 |
+
},
|
| 982 |
+
{
|
| 983 |
+
"epoch": 9.09,
|
| 984 |
+
"eval_accuracy": 0.9925820645613489,
|
| 985 |
+
"eval_f1": 0.9571847507331379,
|
| 986 |
+
"eval_loss": 0.04227477312088013,
|
| 987 |
+
"eval_precision": 0.953111963957951,
|
| 988 |
+
"eval_recall": 0.9612924941097274,
|
| 989 |
+
"eval_runtime": 6.1147,
|
| 990 |
+
"eval_samples_per_second": 531.668,
|
| 991 |
+
"eval_steps_per_second": 0.654,
|
| 992 |
+
"step": 2000
|
| 993 |
+
},
|
| 994 |
+
{
|
| 995 |
+
"epoch": 9.2,
|
| 996 |
+
"eval_accuracy": 0.9925431252677076,
|
| 997 |
+
"eval_f1": 0.9567110441262665,
|
| 998 |
+
"eval_loss": 0.042577099055051804,
|
| 999 |
+
"eval_precision": 0.9520079986668889,
|
| 1000 |
+
"eval_recall": 0.9614607876135981,
|
| 1001 |
+
"eval_runtime": 6.2941,
|
| 1002 |
+
"eval_samples_per_second": 516.519,
|
| 1003 |
+
"eval_steps_per_second": 0.636,
|
| 1004 |
+
"step": 2025
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 9.32,
|
| 1008 |
+
"eval_accuracy": 0.9924652466804252,
|
| 1009 |
+
"eval_f1": 0.9560338330123105,
|
| 1010 |
+
"eval_loss": 0.04250513017177582,
|
| 1011 |
+
"eval_precision": 0.9514919153192198,
|
| 1012 |
+
"eval_recall": 0.9606193200942443,
|
| 1013 |
+
"eval_runtime": 6.0627,
|
| 1014 |
+
"eval_samples_per_second": 536.233,
|
| 1015 |
+
"eval_steps_per_second": 0.66,
|
| 1016 |
+
"step": 2050
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 9.43,
|
| 1020 |
+
"eval_accuracy": 0.9925431252677076,
|
| 1021 |
+
"eval_f1": 0.956463496316142,
|
| 1022 |
+
"eval_loss": 0.0422191396355629,
|
| 1023 |
+
"eval_precision": 0.951682772409197,
|
| 1024 |
+
"eval_recall": 0.9612924941097274,
|
| 1025 |
+
"eval_runtime": 6.244,
|
| 1026 |
+
"eval_samples_per_second": 520.664,
|
| 1027 |
+
"eval_steps_per_second": 0.641,
|
| 1028 |
+
"step": 2075
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 9.55,
|
| 1032 |
+
"eval_accuracy": 0.9924847163272458,
|
| 1033 |
+
"eval_f1": 0.9559537765868363,
|
| 1034 |
+
"eval_loss": 0.0423436276614666,
|
| 1035 |
+
"eval_precision": 0.9513333333333334,
|
| 1036 |
+
"eval_recall": 0.9606193200942443,
|
| 1037 |
+
"eval_runtime": 6.3539,
|
| 1038 |
+
"eval_samples_per_second": 511.652,
|
| 1039 |
+
"eval_steps_per_second": 0.63,
|
| 1040 |
+
"step": 2100
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 9.66,
|
| 1044 |
+
"eval_accuracy": 0.9924652466804252,
|
| 1045 |
+
"eval_f1": 0.9558663428523575,
|
| 1046 |
+
"eval_loss": 0.04235740005970001,
|
| 1047 |
+
"eval_precision": 0.951325220870145,
|
| 1048 |
+
"eval_recall": 0.9604510265903736,
|
| 1049 |
+
"eval_runtime": 6.0977,
|
| 1050 |
+
"eval_samples_per_second": 533.149,
|
| 1051 |
+
"eval_steps_per_second": 0.656,
|
| 1052 |
+
"step": 2125
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 9.77,
|
| 1056 |
+
"eval_accuracy": 0.9924652466804252,
|
| 1057 |
+
"eval_f1": 0.9566164154103852,
|
| 1058 |
+
"eval_loss": 0.042295120656490326,
|
| 1059 |
+
"eval_precision": 0.9521507169056352,
|
| 1060 |
+
"eval_recall": 0.9611242006058566,
|
| 1061 |
+
"eval_runtime": 6.4502,
|
| 1062 |
+
"eval_samples_per_second": 504.016,
|
| 1063 |
+
"eval_steps_per_second": 0.62,
|
| 1064 |
+
"step": 2150
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 9.89,
|
| 1068 |
+
"eval_accuracy": 0.9924847163272458,
|
| 1069 |
+
"eval_f1": 0.9567037936521229,
|
| 1070 |
+
"eval_loss": 0.04226445034146309,
|
| 1071 |
+
"eval_precision": 0.9521586931155193,
|
| 1072 |
+
"eval_recall": 0.9612924941097274,
|
| 1073 |
+
"eval_runtime": 6.2159,
|
| 1074 |
+
"eval_samples_per_second": 523.014,
|
| 1075 |
+
"eval_steps_per_second": 0.644,
|
| 1076 |
+
"step": 2175
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 10.0,
|
| 1080 |
+
"eval_accuracy": 0.9925236556208871,
|
| 1081 |
+
"eval_f1": 0.9570387739720292,
|
| 1082 |
+
"eval_loss": 0.04224075376987457,
|
| 1083 |
+
"eval_precision": 0.952492082013669,
|
| 1084 |
+
"eval_recall": 0.9616290811174689,
|
| 1085 |
+
"eval_runtime": 6.1877,
|
| 1086 |
+
"eval_samples_per_second": 525.397,
|
| 1087 |
+
"eval_steps_per_second": 0.646,
|
| 1088 |
+
"step": 2200
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 10.0,
|
| 1092 |
+
"step": 2200,
|
| 1093 |
+
"total_flos": 4144211411102244.0,
|
| 1094 |
+
"train_loss": 0.021953743980689482,
|
| 1095 |
+
"train_runtime": 1347.1839,
|
| 1096 |
+
"train_samples_per_second": 104.232,
|
| 1097 |
+
"train_steps_per_second": 1.633
|
| 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 |
+
}
|