End of training
Browse files- README.md +1 -1
- all_results.json +40 -0
- eval_results.json +35 -0
- train_results.json +8 -0
- trainer_state.json +1287 -0
README.md
CHANGED
|
@@ -19,7 +19,7 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 19 |
|
| 20 |
This model is a fine-tuned version of [roberta-large](https://huggingface.co/roberta-large) on an unknown dataset.
|
| 21 |
It achieves the following results on the evaluation set:
|
| 22 |
-
- Loss: 1.
|
| 23 |
- Accuracy: 0.4472
|
| 24 |
- Prec: 0.2000
|
| 25 |
- Recall: 0.4472
|
|
|
|
| 19 |
|
| 20 |
This model is a fine-tuned version of [roberta-large](https://huggingface.co/roberta-large) on an unknown dataset.
|
| 21 |
It achieves the following results on the evaluation set:
|
| 22 |
+
- Loss: 1.5366
|
| 23 |
- Accuracy: 0.4472
|
| 24 |
- Prec: 0.2000
|
| 25 |
- Recall: 0.4472
|
all_results.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 25.0,
|
| 3 |
+
"eval_accuracy": 0.4471636546184739,
|
| 4 |
+
"eval_b_acc": 0.14285714285714285,
|
| 5 |
+
"eval_f1": 0.2763410114310333,
|
| 6 |
+
"eval_f1_anger": 0.0,
|
| 7 |
+
"eval_f1_disgust": 0.0,
|
| 8 |
+
"eval_f1_fear": 0.0,
|
| 9 |
+
"eval_f1_joy": 0.0,
|
| 10 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 11 |
+
"eval_f1_sadness": 0.0,
|
| 12 |
+
"eval_f1_surprise": 0.0,
|
| 13 |
+
"eval_loss": 1.5365694761276245,
|
| 14 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 15 |
+
"eval_prec": 0.1999553340117498,
|
| 16 |
+
"eval_prec_anger": 0.0,
|
| 17 |
+
"eval_prec_disgust": 0.0,
|
| 18 |
+
"eval_prec_fear": 0.0,
|
| 19 |
+
"eval_prec_joy": 0.0,
|
| 20 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 21 |
+
"eval_prec_sadness": 0.0,
|
| 22 |
+
"eval_prec_surprise": 0.0,
|
| 23 |
+
"eval_recall": 0.4471636546184739,
|
| 24 |
+
"eval_recall_anger": 0.0,
|
| 25 |
+
"eval_recall_disgust": 0.0,
|
| 26 |
+
"eval_recall_fear": 0.0,
|
| 27 |
+
"eval_recall_joy": 0.0,
|
| 28 |
+
"eval_recall_neutral": 1.0,
|
| 29 |
+
"eval_recall_sadness": 0.0,
|
| 30 |
+
"eval_recall_surprise": 0.0,
|
| 31 |
+
"eval_runtime": 52.8586,
|
| 32 |
+
"eval_samples": 23904,
|
| 33 |
+
"eval_samples_per_second": 452.225,
|
| 34 |
+
"eval_steps_per_second": 14.132,
|
| 35 |
+
"train_loss": 1.4594040881953265,
|
| 36 |
+
"train_runtime": 28719.4719,
|
| 37 |
+
"train_samples": 214113,
|
| 38 |
+
"train_samples_per_second": 186.383,
|
| 39 |
+
"train_steps_per_second": 1.456
|
| 40 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 25.0,
|
| 3 |
+
"eval_accuracy": 0.4471636546184739,
|
| 4 |
+
"eval_b_acc": 0.14285714285714285,
|
| 5 |
+
"eval_f1": 0.2763410114310333,
|
| 6 |
+
"eval_f1_anger": 0.0,
|
| 7 |
+
"eval_f1_disgust": 0.0,
|
| 8 |
+
"eval_f1_fear": 0.0,
|
| 9 |
+
"eval_f1_joy": 0.0,
|
| 10 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 11 |
+
"eval_f1_sadness": 0.0,
|
| 12 |
+
"eval_f1_surprise": 0.0,
|
| 13 |
+
"eval_loss": 1.5365694761276245,
|
| 14 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 15 |
+
"eval_prec": 0.1999553340117498,
|
| 16 |
+
"eval_prec_anger": 0.0,
|
| 17 |
+
"eval_prec_disgust": 0.0,
|
| 18 |
+
"eval_prec_fear": 0.0,
|
| 19 |
+
"eval_prec_joy": 0.0,
|
| 20 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 21 |
+
"eval_prec_sadness": 0.0,
|
| 22 |
+
"eval_prec_surprise": 0.0,
|
| 23 |
+
"eval_recall": 0.4471636546184739,
|
| 24 |
+
"eval_recall_anger": 0.0,
|
| 25 |
+
"eval_recall_disgust": 0.0,
|
| 26 |
+
"eval_recall_fear": 0.0,
|
| 27 |
+
"eval_recall_joy": 0.0,
|
| 28 |
+
"eval_recall_neutral": 1.0,
|
| 29 |
+
"eval_recall_sadness": 0.0,
|
| 30 |
+
"eval_recall_surprise": 0.0,
|
| 31 |
+
"eval_runtime": 52.8586,
|
| 32 |
+
"eval_samples": 23904,
|
| 33 |
+
"eval_samples_per_second": 452.225,
|
| 34 |
+
"eval_steps_per_second": 14.132
|
| 35 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 25.0,
|
| 3 |
+
"train_loss": 1.4594040881953265,
|
| 4 |
+
"train_runtime": 28719.4719,
|
| 5 |
+
"train_samples": 214113,
|
| 6 |
+
"train_samples_per_second": 186.383,
|
| 7 |
+
"train_steps_per_second": 1.456
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1287 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 25.0,
|
| 5 |
+
"eval_steps": 2092,
|
| 6 |
+
"global_step": 41825,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.25,
|
| 13 |
+
"learning_rate": 0.0001988527724665392,
|
| 14 |
+
"loss": 1.1848,
|
| 15 |
+
"step": 419
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"epoch": 0.5,
|
| 19 |
+
"learning_rate": 0.00039913957934990444,
|
| 20 |
+
"loss": 0.8965,
|
| 21 |
+
"step": 838
|
| 22 |
+
},
|
| 23 |
+
{
|
| 24 |
+
"epoch": 0.75,
|
| 25 |
+
"learning_rate": 0.0005994263862332697,
|
| 26 |
+
"loss": 0.8659,
|
| 27 |
+
"step": 1257
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 1.0,
|
| 31 |
+
"learning_rate": 0.0007992351816443595,
|
| 32 |
+
"loss": 0.8381,
|
| 33 |
+
"step": 1676
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"epoch": 1.25,
|
| 37 |
+
"eval_accuracy": 0.4471636546184739,
|
| 38 |
+
"eval_b_acc": 0.14285714285714285,
|
| 39 |
+
"eval_f1": 0.2763410114310333,
|
| 40 |
+
"eval_f1_anger": 0.0,
|
| 41 |
+
"eval_f1_disgust": 0.0,
|
| 42 |
+
"eval_f1_fear": 0.0,
|
| 43 |
+
"eval_f1_joy": 0.0,
|
| 44 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 45 |
+
"eval_f1_sadness": 0.0,
|
| 46 |
+
"eval_f1_surprise": 0.0,
|
| 47 |
+
"eval_loss": 1.541502594947815,
|
| 48 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 49 |
+
"eval_prec": 0.1999553340117498,
|
| 50 |
+
"eval_prec_anger": 0.0,
|
| 51 |
+
"eval_prec_disgust": 0.0,
|
| 52 |
+
"eval_prec_fear": 0.0,
|
| 53 |
+
"eval_prec_joy": 0.0,
|
| 54 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 55 |
+
"eval_prec_sadness": 0.0,
|
| 56 |
+
"eval_prec_surprise": 0.0,
|
| 57 |
+
"eval_recall": 0.4471636546184739,
|
| 58 |
+
"eval_recall_anger": 0.0,
|
| 59 |
+
"eval_recall_disgust": 0.0,
|
| 60 |
+
"eval_recall_fear": 0.0,
|
| 61 |
+
"eval_recall_joy": 0.0,
|
| 62 |
+
"eval_recall_neutral": 1.0,
|
| 63 |
+
"eval_recall_sadness": 0.0,
|
| 64 |
+
"eval_recall_surprise": 0.0,
|
| 65 |
+
"eval_runtime": 53.3613,
|
| 66 |
+
"eval_samples_per_second": 447.965,
|
| 67 |
+
"eval_steps_per_second": 13.999,
|
| 68 |
+
"step": 2092
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 1.25,
|
| 72 |
+
"learning_rate": 0.0009995219885277247,
|
| 73 |
+
"loss": 0.9804,
|
| 74 |
+
"step": 2095
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 1.5,
|
| 78 |
+
"learning_rate": 0.000989479777514912,
|
| 79 |
+
"loss": 1.4905,
|
| 80 |
+
"step": 2514
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 1.75,
|
| 84 |
+
"learning_rate": 0.0009789343870334484,
|
| 85 |
+
"loss": 1.4837,
|
| 86 |
+
"step": 2933
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 2.0,
|
| 90 |
+
"learning_rate": 0.0009683889965519845,
|
| 91 |
+
"loss": 1.4851,
|
| 92 |
+
"step": 3352
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 2.25,
|
| 96 |
+
"learning_rate": 0.0009578436060705208,
|
| 97 |
+
"loss": 1.4866,
|
| 98 |
+
"step": 3771
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 2.5,
|
| 102 |
+
"eval_accuracy": 0.4471636546184739,
|
| 103 |
+
"eval_b_acc": 0.14285714285714285,
|
| 104 |
+
"eval_f1": 0.2763410114310333,
|
| 105 |
+
"eval_f1_anger": 0.0,
|
| 106 |
+
"eval_f1_disgust": 0.0,
|
| 107 |
+
"eval_f1_fear": 0.0,
|
| 108 |
+
"eval_f1_joy": 0.0,
|
| 109 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 110 |
+
"eval_f1_sadness": 0.0,
|
| 111 |
+
"eval_f1_surprise": 0.0,
|
| 112 |
+
"eval_loss": 1.5563576221466064,
|
| 113 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 114 |
+
"eval_prec": 0.1999553340117498,
|
| 115 |
+
"eval_prec_anger": 0.0,
|
| 116 |
+
"eval_prec_disgust": 0.0,
|
| 117 |
+
"eval_prec_fear": 0.0,
|
| 118 |
+
"eval_prec_joy": 0.0,
|
| 119 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 120 |
+
"eval_prec_sadness": 0.0,
|
| 121 |
+
"eval_prec_surprise": 0.0,
|
| 122 |
+
"eval_recall": 0.4471636546184739,
|
| 123 |
+
"eval_recall_anger": 0.0,
|
| 124 |
+
"eval_recall_disgust": 0.0,
|
| 125 |
+
"eval_recall_fear": 0.0,
|
| 126 |
+
"eval_recall_joy": 0.0,
|
| 127 |
+
"eval_recall_neutral": 1.0,
|
| 128 |
+
"eval_recall_sadness": 0.0,
|
| 129 |
+
"eval_recall_surprise": 0.0,
|
| 130 |
+
"eval_runtime": 52.7331,
|
| 131 |
+
"eval_samples_per_second": 453.301,
|
| 132 |
+
"eval_steps_per_second": 14.166,
|
| 133 |
+
"step": 4184
|
| 134 |
+
},
|
| 135 |
+
{
|
| 136 |
+
"epoch": 2.5,
|
| 137 |
+
"learning_rate": 0.000947298215589057,
|
| 138 |
+
"loss": 1.4817,
|
| 139 |
+
"step": 4190
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"epoch": 2.75,
|
| 143 |
+
"learning_rate": 0.0009367528251075931,
|
| 144 |
+
"loss": 1.4846,
|
| 145 |
+
"step": 4609
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"epoch": 3.01,
|
| 149 |
+
"learning_rate": 0.0009262074346261294,
|
| 150 |
+
"loss": 1.4814,
|
| 151 |
+
"step": 5028
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 3.26,
|
| 155 |
+
"learning_rate": 0.0009156872121410415,
|
| 156 |
+
"loss": 1.4818,
|
| 157 |
+
"step": 5447
|
| 158 |
+
},
|
| 159 |
+
{
|
| 160 |
+
"epoch": 3.51,
|
| 161 |
+
"learning_rate": 0.0009051418216595777,
|
| 162 |
+
"loss": 1.4862,
|
| 163 |
+
"step": 5866
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 3.75,
|
| 167 |
+
"eval_accuracy": 0.4471636546184739,
|
| 168 |
+
"eval_b_acc": 0.14285714285714285,
|
| 169 |
+
"eval_f1": 0.2763410114310333,
|
| 170 |
+
"eval_f1_anger": 0.0,
|
| 171 |
+
"eval_f1_disgust": 0.0,
|
| 172 |
+
"eval_f1_fear": 0.0,
|
| 173 |
+
"eval_f1_joy": 0.0,
|
| 174 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 175 |
+
"eval_f1_sadness": 0.0,
|
| 176 |
+
"eval_f1_surprise": 0.0,
|
| 177 |
+
"eval_loss": 1.5699833631515503,
|
| 178 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 179 |
+
"eval_prec": 0.1999553340117498,
|
| 180 |
+
"eval_prec_anger": 0.0,
|
| 181 |
+
"eval_prec_disgust": 0.0,
|
| 182 |
+
"eval_prec_fear": 0.0,
|
| 183 |
+
"eval_prec_joy": 0.0,
|
| 184 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 185 |
+
"eval_prec_sadness": 0.0,
|
| 186 |
+
"eval_prec_surprise": 0.0,
|
| 187 |
+
"eval_recall": 0.4471636546184739,
|
| 188 |
+
"eval_recall_anger": 0.0,
|
| 189 |
+
"eval_recall_disgust": 0.0,
|
| 190 |
+
"eval_recall_fear": 0.0,
|
| 191 |
+
"eval_recall_joy": 0.0,
|
| 192 |
+
"eval_recall_neutral": 1.0,
|
| 193 |
+
"eval_recall_sadness": 0.0,
|
| 194 |
+
"eval_recall_surprise": 0.0,
|
| 195 |
+
"eval_runtime": 52.9507,
|
| 196 |
+
"eval_samples_per_second": 451.439,
|
| 197 |
+
"eval_steps_per_second": 14.107,
|
| 198 |
+
"step": 6276
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 3.76,
|
| 202 |
+
"learning_rate": 0.0008945964311781139,
|
| 203 |
+
"loss": 1.482,
|
| 204 |
+
"step": 6285
|
| 205 |
+
},
|
| 206 |
+
{
|
| 207 |
+
"epoch": 4.01,
|
| 208 |
+
"learning_rate": 0.0008840510406966502,
|
| 209 |
+
"loss": 1.4793,
|
| 210 |
+
"step": 6704
|
| 211 |
+
},
|
| 212 |
+
{
|
| 213 |
+
"epoch": 4.26,
|
| 214 |
+
"learning_rate": 0.0008735056502151864,
|
| 215 |
+
"loss": 1.4753,
|
| 216 |
+
"step": 7123
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 4.51,
|
| 220 |
+
"learning_rate": 0.0008629854277300984,
|
| 221 |
+
"loss": 1.4905,
|
| 222 |
+
"step": 7542
|
| 223 |
+
},
|
| 224 |
+
{
|
| 225 |
+
"epoch": 4.76,
|
| 226 |
+
"learning_rate": 0.0008524400372486346,
|
| 227 |
+
"loss": 1.4762,
|
| 228 |
+
"step": 7961
|
| 229 |
+
},
|
| 230 |
+
{
|
| 231 |
+
"epoch": 5.0,
|
| 232 |
+
"eval_accuracy": 0.4471636546184739,
|
| 233 |
+
"eval_b_acc": 0.14285714285714285,
|
| 234 |
+
"eval_f1": 0.2763410114310333,
|
| 235 |
+
"eval_f1_anger": 0.0,
|
| 236 |
+
"eval_f1_disgust": 0.0,
|
| 237 |
+
"eval_f1_fear": 0.0,
|
| 238 |
+
"eval_f1_joy": 0.0,
|
| 239 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 240 |
+
"eval_f1_sadness": 0.0,
|
| 241 |
+
"eval_f1_surprise": 0.0,
|
| 242 |
+
"eval_loss": 1.5391422510147095,
|
| 243 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 244 |
+
"eval_prec": 0.1999553340117498,
|
| 245 |
+
"eval_prec_anger": 0.0,
|
| 246 |
+
"eval_prec_disgust": 0.0,
|
| 247 |
+
"eval_prec_fear": 0.0,
|
| 248 |
+
"eval_prec_joy": 0.0,
|
| 249 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 250 |
+
"eval_prec_sadness": 0.0,
|
| 251 |
+
"eval_prec_surprise": 0.0,
|
| 252 |
+
"eval_recall": 0.4471636546184739,
|
| 253 |
+
"eval_recall_anger": 0.0,
|
| 254 |
+
"eval_recall_disgust": 0.0,
|
| 255 |
+
"eval_recall_fear": 0.0,
|
| 256 |
+
"eval_recall_joy": 0.0,
|
| 257 |
+
"eval_recall_neutral": 1.0,
|
| 258 |
+
"eval_recall_sadness": 0.0,
|
| 259 |
+
"eval_recall_surprise": 0.0,
|
| 260 |
+
"eval_runtime": 52.76,
|
| 261 |
+
"eval_samples_per_second": 453.07,
|
| 262 |
+
"eval_steps_per_second": 14.158,
|
| 263 |
+
"step": 8368
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 5.01,
|
| 267 |
+
"learning_rate": 0.0008418946467671709,
|
| 268 |
+
"loss": 1.4794,
|
| 269 |
+
"step": 8380
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"epoch": 5.26,
|
| 273 |
+
"learning_rate": 0.0008313995922784587,
|
| 274 |
+
"loss": 1.48,
|
| 275 |
+
"step": 8799
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 5.51,
|
| 279 |
+
"learning_rate": 0.0008209045377897466,
|
| 280 |
+
"loss": 1.4795,
|
| 281 |
+
"step": 9218
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 5.76,
|
| 285 |
+
"learning_rate": 0.0008104094833010344,
|
| 286 |
+
"loss": 1.4815,
|
| 287 |
+
"step": 9637
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"epoch": 6.01,
|
| 291 |
+
"learning_rate": 0.0007998640928195707,
|
| 292 |
+
"loss": 1.4765,
|
| 293 |
+
"step": 10056
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 6.25,
|
| 297 |
+
"eval_accuracy": 0.4471636546184739,
|
| 298 |
+
"eval_b_acc": 0.14285714285714285,
|
| 299 |
+
"eval_f1": 0.2763410114310333,
|
| 300 |
+
"eval_f1_anger": 0.0,
|
| 301 |
+
"eval_f1_disgust": 0.0,
|
| 302 |
+
"eval_f1_fear": 0.0,
|
| 303 |
+
"eval_f1_joy": 0.0,
|
| 304 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 305 |
+
"eval_f1_sadness": 0.0,
|
| 306 |
+
"eval_f1_surprise": 0.0,
|
| 307 |
+
"eval_loss": 1.5565674304962158,
|
| 308 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 309 |
+
"eval_prec": 0.1999553340117498,
|
| 310 |
+
"eval_prec_anger": 0.0,
|
| 311 |
+
"eval_prec_disgust": 0.0,
|
| 312 |
+
"eval_prec_fear": 0.0,
|
| 313 |
+
"eval_prec_joy": 0.0,
|
| 314 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 315 |
+
"eval_prec_sadness": 0.0,
|
| 316 |
+
"eval_prec_surprise": 0.0,
|
| 317 |
+
"eval_recall": 0.4471636546184739,
|
| 318 |
+
"eval_recall_anger": 0.0,
|
| 319 |
+
"eval_recall_disgust": 0.0,
|
| 320 |
+
"eval_recall_fear": 0.0,
|
| 321 |
+
"eval_recall_joy": 0.0,
|
| 322 |
+
"eval_recall_neutral": 1.0,
|
| 323 |
+
"eval_recall_sadness": 0.0,
|
| 324 |
+
"eval_recall_surprise": 0.0,
|
| 325 |
+
"eval_runtime": 52.8225,
|
| 326 |
+
"eval_samples_per_second": 452.534,
|
| 327 |
+
"eval_steps_per_second": 14.142,
|
| 328 |
+
"step": 10460
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 6.26,
|
| 332 |
+
"learning_rate": 0.0007893187023381068,
|
| 333 |
+
"loss": 1.4785,
|
| 334 |
+
"step": 10475
|
| 335 |
+
},
|
| 336 |
+
{
|
| 337 |
+
"epoch": 6.51,
|
| 338 |
+
"learning_rate": 0.0007787733118566431,
|
| 339 |
+
"loss": 1.4841,
|
| 340 |
+
"step": 10894
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 6.76,
|
| 344 |
+
"learning_rate": 0.0007682279213751794,
|
| 345 |
+
"loss": 1.4815,
|
| 346 |
+
"step": 11313
|
| 347 |
+
},
|
| 348 |
+
{
|
| 349 |
+
"epoch": 7.01,
|
| 350 |
+
"learning_rate": 0.0007577328668864671,
|
| 351 |
+
"loss": 1.4808,
|
| 352 |
+
"step": 11732
|
| 353 |
+
},
|
| 354 |
+
{
|
| 355 |
+
"epoch": 7.26,
|
| 356 |
+
"learning_rate": 0.0007471874764050034,
|
| 357 |
+
"loss": 1.4848,
|
| 358 |
+
"step": 12151
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"epoch": 7.5,
|
| 362 |
+
"eval_accuracy": 0.4471636546184739,
|
| 363 |
+
"eval_b_acc": 0.14285714285714285,
|
| 364 |
+
"eval_f1": 0.2763410114310333,
|
| 365 |
+
"eval_f1_anger": 0.0,
|
| 366 |
+
"eval_f1_disgust": 0.0,
|
| 367 |
+
"eval_f1_fear": 0.0,
|
| 368 |
+
"eval_f1_joy": 0.0,
|
| 369 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 370 |
+
"eval_f1_sadness": 0.0,
|
| 371 |
+
"eval_f1_surprise": 0.0,
|
| 372 |
+
"eval_loss": 1.5410676002502441,
|
| 373 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 374 |
+
"eval_prec": 0.1999553340117498,
|
| 375 |
+
"eval_prec_anger": 0.0,
|
| 376 |
+
"eval_prec_disgust": 0.0,
|
| 377 |
+
"eval_prec_fear": 0.0,
|
| 378 |
+
"eval_prec_joy": 0.0,
|
| 379 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 380 |
+
"eval_prec_sadness": 0.0,
|
| 381 |
+
"eval_prec_surprise": 0.0,
|
| 382 |
+
"eval_recall": 0.4471636546184739,
|
| 383 |
+
"eval_recall_anger": 0.0,
|
| 384 |
+
"eval_recall_disgust": 0.0,
|
| 385 |
+
"eval_recall_fear": 0.0,
|
| 386 |
+
"eval_recall_joy": 0.0,
|
| 387 |
+
"eval_recall_neutral": 1.0,
|
| 388 |
+
"eval_recall_sadness": 0.0,
|
| 389 |
+
"eval_recall_surprise": 0.0,
|
| 390 |
+
"eval_runtime": 53.1004,
|
| 391 |
+
"eval_samples_per_second": 450.166,
|
| 392 |
+
"eval_steps_per_second": 14.068,
|
| 393 |
+
"step": 12552
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 7.51,
|
| 397 |
+
"learning_rate": 0.0007366420859235397,
|
| 398 |
+
"loss": 1.4779,
|
| 399 |
+
"step": 12570
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 7.76,
|
| 403 |
+
"learning_rate": 0.0007260966954420759,
|
| 404 |
+
"loss": 1.4833,
|
| 405 |
+
"step": 12989
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 8.01,
|
| 409 |
+
"learning_rate": 0.0007155764729569879,
|
| 410 |
+
"loss": 1.4759,
|
| 411 |
+
"step": 13408
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 8.26,
|
| 415 |
+
"learning_rate": 0.0007050310824755242,
|
| 416 |
+
"loss": 1.4763,
|
| 417 |
+
"step": 13827
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 8.52,
|
| 421 |
+
"learning_rate": 0.0006945611959831878,
|
| 422 |
+
"loss": 1.4782,
|
| 423 |
+
"step": 14246
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"epoch": 8.75,
|
| 427 |
+
"eval_accuracy": 0.4471636546184739,
|
| 428 |
+
"eval_b_acc": 0.14285714285714285,
|
| 429 |
+
"eval_f1": 0.2763410114310333,
|
| 430 |
+
"eval_f1_anger": 0.0,
|
| 431 |
+
"eval_f1_disgust": 0.0,
|
| 432 |
+
"eval_f1_fear": 0.0,
|
| 433 |
+
"eval_f1_joy": 0.0,
|
| 434 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 435 |
+
"eval_f1_sadness": 0.0,
|
| 436 |
+
"eval_f1_surprise": 0.0,
|
| 437 |
+
"eval_loss": 1.5548430681228638,
|
| 438 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 439 |
+
"eval_prec": 0.1999553340117498,
|
| 440 |
+
"eval_prec_anger": 0.0,
|
| 441 |
+
"eval_prec_disgust": 0.0,
|
| 442 |
+
"eval_prec_fear": 0.0,
|
| 443 |
+
"eval_prec_joy": 0.0,
|
| 444 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 445 |
+
"eval_prec_sadness": 0.0,
|
| 446 |
+
"eval_prec_surprise": 0.0,
|
| 447 |
+
"eval_recall": 0.4471636546184739,
|
| 448 |
+
"eval_recall_anger": 0.0,
|
| 449 |
+
"eval_recall_disgust": 0.0,
|
| 450 |
+
"eval_recall_fear": 0.0,
|
| 451 |
+
"eval_recall_joy": 0.0,
|
| 452 |
+
"eval_recall_neutral": 1.0,
|
| 453 |
+
"eval_recall_sadness": 0.0,
|
| 454 |
+
"eval_recall_surprise": 0.0,
|
| 455 |
+
"eval_runtime": 52.8918,
|
| 456 |
+
"eval_samples_per_second": 451.941,
|
| 457 |
+
"eval_steps_per_second": 14.123,
|
| 458 |
+
"step": 14644
|
| 459 |
+
},
|
| 460 |
+
{
|
| 461 |
+
"epoch": 8.77,
|
| 462 |
+
"learning_rate": 0.0006840158055017241,
|
| 463 |
+
"loss": 1.482,
|
| 464 |
+
"step": 14665
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 9.02,
|
| 468 |
+
"learning_rate": 0.0006734704150202603,
|
| 469 |
+
"loss": 1.4789,
|
| 470 |
+
"step": 15084
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 9.27,
|
| 474 |
+
"learning_rate": 0.0006629753605315481,
|
| 475 |
+
"loss": 1.4715,
|
| 476 |
+
"step": 15503
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 9.52,
|
| 480 |
+
"learning_rate": 0.0006524299700500843,
|
| 481 |
+
"loss": 1.4967,
|
| 482 |
+
"step": 15922
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 9.77,
|
| 486 |
+
"learning_rate": 0.0006418845795686205,
|
| 487 |
+
"loss": 1.4943,
|
| 488 |
+
"step": 16341
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 10.0,
|
| 492 |
+
"eval_accuracy": 0.4471636546184739,
|
| 493 |
+
"eval_b_acc": 0.14285714285714285,
|
| 494 |
+
"eval_f1": 0.2763410114310333,
|
| 495 |
+
"eval_f1_anger": 0.0,
|
| 496 |
+
"eval_f1_disgust": 0.0,
|
| 497 |
+
"eval_f1_fear": 0.0,
|
| 498 |
+
"eval_f1_joy": 0.0,
|
| 499 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 500 |
+
"eval_f1_sadness": 0.0,
|
| 501 |
+
"eval_f1_surprise": 0.0,
|
| 502 |
+
"eval_loss": 1.6114758253097534,
|
| 503 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 504 |
+
"eval_prec": 0.1999553340117498,
|
| 505 |
+
"eval_prec_anger": 0.0,
|
| 506 |
+
"eval_prec_disgust": 0.0,
|
| 507 |
+
"eval_prec_fear": 0.0,
|
| 508 |
+
"eval_prec_joy": 0.0,
|
| 509 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 510 |
+
"eval_prec_sadness": 0.0,
|
| 511 |
+
"eval_prec_surprise": 0.0,
|
| 512 |
+
"eval_recall": 0.4471636546184739,
|
| 513 |
+
"eval_recall_anger": 0.0,
|
| 514 |
+
"eval_recall_disgust": 0.0,
|
| 515 |
+
"eval_recall_fear": 0.0,
|
| 516 |
+
"eval_recall_joy": 0.0,
|
| 517 |
+
"eval_recall_neutral": 1.0,
|
| 518 |
+
"eval_recall_sadness": 0.0,
|
| 519 |
+
"eval_recall_surprise": 0.0,
|
| 520 |
+
"eval_runtime": 53.2358,
|
| 521 |
+
"eval_samples_per_second": 449.021,
|
| 522 |
+
"eval_steps_per_second": 14.032,
|
| 523 |
+
"step": 16736
|
| 524 |
+
},
|
| 525 |
+
{
|
| 526 |
+
"epoch": 10.02,
|
| 527 |
+
"learning_rate": 0.0006313391890871568,
|
| 528 |
+
"loss": 1.4874,
|
| 529 |
+
"step": 16760
|
| 530 |
+
},
|
| 531 |
+
{
|
| 532 |
+
"epoch": 10.27,
|
| 533 |
+
"learning_rate": 0.000620793798605693,
|
| 534 |
+
"loss": 1.4877,
|
| 535 |
+
"step": 17179
|
| 536 |
+
},
|
| 537 |
+
{
|
| 538 |
+
"epoch": 10.52,
|
| 539 |
+
"learning_rate": 0.0006102484081242293,
|
| 540 |
+
"loss": 1.4833,
|
| 541 |
+
"step": 17598
|
| 542 |
+
},
|
| 543 |
+
{
|
| 544 |
+
"epoch": 10.77,
|
| 545 |
+
"learning_rate": 0.0005997281856391413,
|
| 546 |
+
"loss": 1.4796,
|
| 547 |
+
"step": 18017
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 11.02,
|
| 551 |
+
"learning_rate": 0.0005891827951576775,
|
| 552 |
+
"loss": 1.4801,
|
| 553 |
+
"step": 18436
|
| 554 |
+
},
|
| 555 |
+
{
|
| 556 |
+
"epoch": 11.25,
|
| 557 |
+
"eval_accuracy": 0.4471636546184739,
|
| 558 |
+
"eval_b_acc": 0.14285714285714285,
|
| 559 |
+
"eval_f1": 0.2763410114310333,
|
| 560 |
+
"eval_f1_anger": 0.0,
|
| 561 |
+
"eval_f1_disgust": 0.0,
|
| 562 |
+
"eval_f1_fear": 0.0,
|
| 563 |
+
"eval_f1_joy": 0.0,
|
| 564 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 565 |
+
"eval_f1_sadness": 0.0,
|
| 566 |
+
"eval_f1_surprise": 0.0,
|
| 567 |
+
"eval_loss": 1.5423938035964966,
|
| 568 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 569 |
+
"eval_prec": 0.1999553340117498,
|
| 570 |
+
"eval_prec_anger": 0.0,
|
| 571 |
+
"eval_prec_disgust": 0.0,
|
| 572 |
+
"eval_prec_fear": 0.0,
|
| 573 |
+
"eval_prec_joy": 0.0,
|
| 574 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 575 |
+
"eval_prec_sadness": 0.0,
|
| 576 |
+
"eval_prec_surprise": 0.0,
|
| 577 |
+
"eval_recall": 0.4471636546184739,
|
| 578 |
+
"eval_recall_anger": 0.0,
|
| 579 |
+
"eval_recall_disgust": 0.0,
|
| 580 |
+
"eval_recall_fear": 0.0,
|
| 581 |
+
"eval_recall_joy": 0.0,
|
| 582 |
+
"eval_recall_neutral": 1.0,
|
| 583 |
+
"eval_recall_sadness": 0.0,
|
| 584 |
+
"eval_recall_surprise": 0.0,
|
| 585 |
+
"eval_runtime": 52.871,
|
| 586 |
+
"eval_samples_per_second": 452.119,
|
| 587 |
+
"eval_steps_per_second": 14.129,
|
| 588 |
+
"step": 18828
|
| 589 |
+
},
|
| 590 |
+
{
|
| 591 |
+
"epoch": 11.27,
|
| 592 |
+
"learning_rate": 0.0005786374046762138,
|
| 593 |
+
"loss": 1.4804,
|
| 594 |
+
"step": 18855
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"epoch": 11.52,
|
| 598 |
+
"learning_rate": 0.0005680920141947499,
|
| 599 |
+
"loss": 1.4836,
|
| 600 |
+
"step": 19274
|
| 601 |
+
},
|
| 602 |
+
{
|
| 603 |
+
"epoch": 11.77,
|
| 604 |
+
"learning_rate": 0.0005575466237132862,
|
| 605 |
+
"loss": 1.4839,
|
| 606 |
+
"step": 19693
|
| 607 |
+
},
|
| 608 |
+
{
|
| 609 |
+
"epoch": 12.02,
|
| 610 |
+
"learning_rate": 0.0005470012332318224,
|
| 611 |
+
"loss": 1.4826,
|
| 612 |
+
"step": 20112
|
| 613 |
+
},
|
| 614 |
+
{
|
| 615 |
+
"epoch": 12.27,
|
| 616 |
+
"learning_rate": 0.0005365061787431102,
|
| 617 |
+
"loss": 1.4946,
|
| 618 |
+
"step": 20531
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 12.5,
|
| 622 |
+
"eval_accuracy": 0.4471636546184739,
|
| 623 |
+
"eval_b_acc": 0.14285714285714285,
|
| 624 |
+
"eval_f1": 0.2763410114310333,
|
| 625 |
+
"eval_f1_anger": 0.0,
|
| 626 |
+
"eval_f1_disgust": 0.0,
|
| 627 |
+
"eval_f1_fear": 0.0,
|
| 628 |
+
"eval_f1_joy": 0.0,
|
| 629 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 630 |
+
"eval_f1_sadness": 0.0,
|
| 631 |
+
"eval_f1_surprise": 0.0,
|
| 632 |
+
"eval_loss": 1.5636779069900513,
|
| 633 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 634 |
+
"eval_prec": 0.1999553340117498,
|
| 635 |
+
"eval_prec_anger": 0.0,
|
| 636 |
+
"eval_prec_disgust": 0.0,
|
| 637 |
+
"eval_prec_fear": 0.0,
|
| 638 |
+
"eval_prec_joy": 0.0,
|
| 639 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 640 |
+
"eval_prec_sadness": 0.0,
|
| 641 |
+
"eval_prec_surprise": 0.0,
|
| 642 |
+
"eval_recall": 0.4471636546184739,
|
| 643 |
+
"eval_recall_anger": 0.0,
|
| 644 |
+
"eval_recall_disgust": 0.0,
|
| 645 |
+
"eval_recall_fear": 0.0,
|
| 646 |
+
"eval_recall_joy": 0.0,
|
| 647 |
+
"eval_recall_neutral": 1.0,
|
| 648 |
+
"eval_recall_sadness": 0.0,
|
| 649 |
+
"eval_recall_surprise": 0.0,
|
| 650 |
+
"eval_runtime": 52.887,
|
| 651 |
+
"eval_samples_per_second": 451.982,
|
| 652 |
+
"eval_steps_per_second": 14.124,
|
| 653 |
+
"step": 20920
|
| 654 |
+
},
|
| 655 |
+
{
|
| 656 |
+
"epoch": 12.52,
|
| 657 |
+
"learning_rate": 0.0005259859562580223,
|
| 658 |
+
"loss": 1.4926,
|
| 659 |
+
"step": 20950
|
| 660 |
+
},
|
| 661 |
+
{
|
| 662 |
+
"epoch": 12.77,
|
| 663 |
+
"learning_rate": 0.0005154405657765586,
|
| 664 |
+
"loss": 1.4868,
|
| 665 |
+
"step": 21369
|
| 666 |
+
},
|
| 667 |
+
{
|
| 668 |
+
"epoch": 13.02,
|
| 669 |
+
"learning_rate": 0.0005048951752950947,
|
| 670 |
+
"loss": 1.4776,
|
| 671 |
+
"step": 21788
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 13.27,
|
| 675 |
+
"learning_rate": 0.000494349784813631,
|
| 676 |
+
"loss": 1.4827,
|
| 677 |
+
"step": 22207
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 13.52,
|
| 681 |
+
"learning_rate": 0.00048380439433216724,
|
| 682 |
+
"loss": 1.4867,
|
| 683 |
+
"step": 22626
|
| 684 |
+
},
|
| 685 |
+
{
|
| 686 |
+
"epoch": 13.75,
|
| 687 |
+
"eval_accuracy": 0.4471636546184739,
|
| 688 |
+
"eval_b_acc": 0.14285714285714285,
|
| 689 |
+
"eval_f1": 0.2763410114310333,
|
| 690 |
+
"eval_f1_anger": 0.0,
|
| 691 |
+
"eval_f1_disgust": 0.0,
|
| 692 |
+
"eval_f1_fear": 0.0,
|
| 693 |
+
"eval_f1_joy": 0.0,
|
| 694 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 695 |
+
"eval_f1_sadness": 0.0,
|
| 696 |
+
"eval_f1_surprise": 0.0,
|
| 697 |
+
"eval_loss": 1.5492433309555054,
|
| 698 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 699 |
+
"eval_prec": 0.1999553340117498,
|
| 700 |
+
"eval_prec_anger": 0.0,
|
| 701 |
+
"eval_prec_disgust": 0.0,
|
| 702 |
+
"eval_prec_fear": 0.0,
|
| 703 |
+
"eval_prec_joy": 0.0,
|
| 704 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 705 |
+
"eval_prec_sadness": 0.0,
|
| 706 |
+
"eval_prec_surprise": 0.0,
|
| 707 |
+
"eval_recall": 0.4471636546184739,
|
| 708 |
+
"eval_recall_anger": 0.0,
|
| 709 |
+
"eval_recall_disgust": 0.0,
|
| 710 |
+
"eval_recall_fear": 0.0,
|
| 711 |
+
"eval_recall_joy": 0.0,
|
| 712 |
+
"eval_recall_neutral": 1.0,
|
| 713 |
+
"eval_recall_sadness": 0.0,
|
| 714 |
+
"eval_recall_surprise": 0.0,
|
| 715 |
+
"eval_runtime": 53.2702,
|
| 716 |
+
"eval_samples_per_second": 448.731,
|
| 717 |
+
"eval_steps_per_second": 14.023,
|
| 718 |
+
"step": 23012
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 13.77,
|
| 722 |
+
"learning_rate": 0.00047325900385070346,
|
| 723 |
+
"loss": 1.4902,
|
| 724 |
+
"step": 23045
|
| 725 |
+
},
|
| 726 |
+
{
|
| 727 |
+
"epoch": 14.03,
|
| 728 |
+
"learning_rate": 0.0004627136133692397,
|
| 729 |
+
"loss": 1.4825,
|
| 730 |
+
"step": 23464
|
| 731 |
+
},
|
| 732 |
+
{
|
| 733 |
+
"epoch": 14.28,
|
| 734 |
+
"learning_rate": 0.0004521682228877759,
|
| 735 |
+
"loss": 1.4823,
|
| 736 |
+
"step": 23883
|
| 737 |
+
},
|
| 738 |
+
{
|
| 739 |
+
"epoch": 14.53,
|
| 740 |
+
"learning_rate": 0.00044162283240631214,
|
| 741 |
+
"loss": 1.4933,
|
| 742 |
+
"step": 24302
|
| 743 |
+
},
|
| 744 |
+
{
|
| 745 |
+
"epoch": 14.78,
|
| 746 |
+
"learning_rate": 0.0004311026099212242,
|
| 747 |
+
"loss": 1.4957,
|
| 748 |
+
"step": 24721
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 15.01,
|
| 752 |
+
"eval_accuracy": 0.4471636546184739,
|
| 753 |
+
"eval_b_acc": 0.14285714285714285,
|
| 754 |
+
"eval_f1": 0.2763410114310333,
|
| 755 |
+
"eval_f1_anger": 0.0,
|
| 756 |
+
"eval_f1_disgust": 0.0,
|
| 757 |
+
"eval_f1_fear": 0.0,
|
| 758 |
+
"eval_f1_joy": 0.0,
|
| 759 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 760 |
+
"eval_f1_sadness": 0.0,
|
| 761 |
+
"eval_f1_surprise": 0.0,
|
| 762 |
+
"eval_loss": 1.5811705589294434,
|
| 763 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 764 |
+
"eval_prec": 0.1999553340117498,
|
| 765 |
+
"eval_prec_anger": 0.0,
|
| 766 |
+
"eval_prec_disgust": 0.0,
|
| 767 |
+
"eval_prec_fear": 0.0,
|
| 768 |
+
"eval_prec_joy": 0.0,
|
| 769 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 770 |
+
"eval_prec_sadness": 0.0,
|
| 771 |
+
"eval_prec_surprise": 0.0,
|
| 772 |
+
"eval_recall": 0.4471636546184739,
|
| 773 |
+
"eval_recall_anger": 0.0,
|
| 774 |
+
"eval_recall_disgust": 0.0,
|
| 775 |
+
"eval_recall_fear": 0.0,
|
| 776 |
+
"eval_recall_joy": 0.0,
|
| 777 |
+
"eval_recall_neutral": 1.0,
|
| 778 |
+
"eval_recall_sadness": 0.0,
|
| 779 |
+
"eval_recall_surprise": 0.0,
|
| 780 |
+
"eval_runtime": 52.7987,
|
| 781 |
+
"eval_samples_per_second": 452.739,
|
| 782 |
+
"eval_steps_per_second": 14.148,
|
| 783 |
+
"step": 25104
|
| 784 |
+
},
|
| 785 |
+
{
|
| 786 |
+
"epoch": 15.03,
|
| 787 |
+
"learning_rate": 0.00042055721943976037,
|
| 788 |
+
"loss": 1.4875,
|
| 789 |
+
"step": 25140
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"epoch": 15.28,
|
| 793 |
+
"learning_rate": 0.00041001182895829665,
|
| 794 |
+
"loss": 1.4866,
|
| 795 |
+
"step": 25559
|
| 796 |
+
},
|
| 797 |
+
{
|
| 798 |
+
"epoch": 15.53,
|
| 799 |
+
"learning_rate": 0.00039946643847683287,
|
| 800 |
+
"loss": 1.4879,
|
| 801 |
+
"step": 25978
|
| 802 |
+
},
|
| 803 |
+
{
|
| 804 |
+
"epoch": 15.78,
|
| 805 |
+
"learning_rate": 0.0003889210479953691,
|
| 806 |
+
"loss": 1.4856,
|
| 807 |
+
"step": 26397
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 16.03,
|
| 811 |
+
"learning_rate": 0.00037840082551028115,
|
| 812 |
+
"loss": 1.4913,
|
| 813 |
+
"step": 26816
|
| 814 |
+
},
|
| 815 |
+
{
|
| 816 |
+
"epoch": 16.26,
|
| 817 |
+
"eval_accuracy": 0.4471636546184739,
|
| 818 |
+
"eval_b_acc": 0.14285714285714285,
|
| 819 |
+
"eval_f1": 0.2763410114310333,
|
| 820 |
+
"eval_f1_anger": 0.0,
|
| 821 |
+
"eval_f1_disgust": 0.0,
|
| 822 |
+
"eval_f1_fear": 0.0,
|
| 823 |
+
"eval_f1_joy": 0.0,
|
| 824 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 825 |
+
"eval_f1_sadness": 0.0,
|
| 826 |
+
"eval_f1_surprise": 0.0,
|
| 827 |
+
"eval_loss": 1.5424742698669434,
|
| 828 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 829 |
+
"eval_prec": 0.1999553340117498,
|
| 830 |
+
"eval_prec_anger": 0.0,
|
| 831 |
+
"eval_prec_disgust": 0.0,
|
| 832 |
+
"eval_prec_fear": 0.0,
|
| 833 |
+
"eval_prec_joy": 0.0,
|
| 834 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 835 |
+
"eval_prec_sadness": 0.0,
|
| 836 |
+
"eval_prec_surprise": 0.0,
|
| 837 |
+
"eval_recall": 0.4471636546184739,
|
| 838 |
+
"eval_recall_anger": 0.0,
|
| 839 |
+
"eval_recall_disgust": 0.0,
|
| 840 |
+
"eval_recall_fear": 0.0,
|
| 841 |
+
"eval_recall_joy": 0.0,
|
| 842 |
+
"eval_recall_neutral": 1.0,
|
| 843 |
+
"eval_recall_sadness": 0.0,
|
| 844 |
+
"eval_recall_surprise": 0.0,
|
| 845 |
+
"eval_runtime": 53.2257,
|
| 846 |
+
"eval_samples_per_second": 449.106,
|
| 847 |
+
"eval_steps_per_second": 14.035,
|
| 848 |
+
"step": 27196
|
| 849 |
+
},
|
| 850 |
+
{
|
| 851 |
+
"epoch": 16.28,
|
| 852 |
+
"learning_rate": 0.000367905771021569,
|
| 853 |
+
"loss": 1.4935,
|
| 854 |
+
"step": 27235
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 16.53,
|
| 858 |
+
"learning_rate": 0.00035738554853648104,
|
| 859 |
+
"loss": 1.5047,
|
| 860 |
+
"step": 27654
|
| 861 |
+
},
|
| 862 |
+
{
|
| 863 |
+
"epoch": 16.78,
|
| 864 |
+
"learning_rate": 0.0003468401580550172,
|
| 865 |
+
"loss": 1.4959,
|
| 866 |
+
"step": 28073
|
| 867 |
+
},
|
| 868 |
+
{
|
| 869 |
+
"epoch": 17.03,
|
| 870 |
+
"learning_rate": 0.0003362947675735535,
|
| 871 |
+
"loss": 1.5074,
|
| 872 |
+
"step": 28492
|
| 873 |
+
},
|
| 874 |
+
{
|
| 875 |
+
"epoch": 17.28,
|
| 876 |
+
"learning_rate": 0.00032574937709208967,
|
| 877 |
+
"loss": 1.5007,
|
| 878 |
+
"step": 28911
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 17.51,
|
| 882 |
+
"eval_accuracy": 0.4471636546184739,
|
| 883 |
+
"eval_b_acc": 0.14285714285714285,
|
| 884 |
+
"eval_f1": 0.2763410114310333,
|
| 885 |
+
"eval_f1_anger": 0.0,
|
| 886 |
+
"eval_f1_disgust": 0.0,
|
| 887 |
+
"eval_f1_fear": 0.0,
|
| 888 |
+
"eval_f1_joy": 0.0,
|
| 889 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 890 |
+
"eval_f1_sadness": 0.0,
|
| 891 |
+
"eval_f1_surprise": 0.0,
|
| 892 |
+
"eval_loss": 1.5446112155914307,
|
| 893 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 894 |
+
"eval_prec": 0.1999553340117498,
|
| 895 |
+
"eval_prec_anger": 0.0,
|
| 896 |
+
"eval_prec_disgust": 0.0,
|
| 897 |
+
"eval_prec_fear": 0.0,
|
| 898 |
+
"eval_prec_joy": 0.0,
|
| 899 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 900 |
+
"eval_prec_sadness": 0.0,
|
| 901 |
+
"eval_prec_surprise": 0.0,
|
| 902 |
+
"eval_recall": 0.4471636546184739,
|
| 903 |
+
"eval_recall_anger": 0.0,
|
| 904 |
+
"eval_recall_disgust": 0.0,
|
| 905 |
+
"eval_recall_fear": 0.0,
|
| 906 |
+
"eval_recall_joy": 0.0,
|
| 907 |
+
"eval_recall_neutral": 1.0,
|
| 908 |
+
"eval_recall_sadness": 0.0,
|
| 909 |
+
"eval_recall_surprise": 0.0,
|
| 910 |
+
"eval_runtime": 52.751,
|
| 911 |
+
"eval_samples_per_second": 453.148,
|
| 912 |
+
"eval_steps_per_second": 14.161,
|
| 913 |
+
"step": 29288
|
| 914 |
+
},
|
| 915 |
+
{
|
| 916 |
+
"epoch": 17.53,
|
| 917 |
+
"learning_rate": 0.00031520398661062594,
|
| 918 |
+
"loss": 1.5003,
|
| 919 |
+
"step": 29330
|
| 920 |
+
},
|
| 921 |
+
{
|
| 922 |
+
"epoch": 17.78,
|
| 923 |
+
"learning_rate": 0.00030465859612916217,
|
| 924 |
+
"loss": 1.5044,
|
| 925 |
+
"step": 29749
|
| 926 |
+
},
|
| 927 |
+
{
|
| 928 |
+
"epoch": 18.03,
|
| 929 |
+
"learning_rate": 0.0002941132056476984,
|
| 930 |
+
"loss": 1.4958,
|
| 931 |
+
"step": 30168
|
| 932 |
+
},
|
| 933 |
+
{
|
| 934 |
+
"epoch": 18.28,
|
| 935 |
+
"learning_rate": 0.0002835678151662346,
|
| 936 |
+
"loss": 1.5033,
|
| 937 |
+
"step": 30587
|
| 938 |
+
},
|
| 939 |
+
{
|
| 940 |
+
"epoch": 18.53,
|
| 941 |
+
"learning_rate": 0.00027302242468477085,
|
| 942 |
+
"loss": 1.4919,
|
| 943 |
+
"step": 31006
|
| 944 |
+
},
|
| 945 |
+
{
|
| 946 |
+
"epoch": 18.76,
|
| 947 |
+
"eval_accuracy": 0.4471636546184739,
|
| 948 |
+
"eval_b_acc": 0.14285714285714285,
|
| 949 |
+
"eval_f1": 0.2763410114310333,
|
| 950 |
+
"eval_f1_anger": 0.0,
|
| 951 |
+
"eval_f1_disgust": 0.0,
|
| 952 |
+
"eval_f1_fear": 0.0,
|
| 953 |
+
"eval_f1_joy": 0.0,
|
| 954 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 955 |
+
"eval_f1_sadness": 0.0,
|
| 956 |
+
"eval_f1_surprise": 0.0,
|
| 957 |
+
"eval_loss": 1.5616443157196045,
|
| 958 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 959 |
+
"eval_prec": 0.1999553340117498,
|
| 960 |
+
"eval_prec_anger": 0.0,
|
| 961 |
+
"eval_prec_disgust": 0.0,
|
| 962 |
+
"eval_prec_fear": 0.0,
|
| 963 |
+
"eval_prec_joy": 0.0,
|
| 964 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 965 |
+
"eval_prec_sadness": 0.0,
|
| 966 |
+
"eval_prec_surprise": 0.0,
|
| 967 |
+
"eval_recall": 0.4471636546184739,
|
| 968 |
+
"eval_recall_anger": 0.0,
|
| 969 |
+
"eval_recall_disgust": 0.0,
|
| 970 |
+
"eval_recall_fear": 0.0,
|
| 971 |
+
"eval_recall_joy": 0.0,
|
| 972 |
+
"eval_recall_neutral": 1.0,
|
| 973 |
+
"eval_recall_sadness": 0.0,
|
| 974 |
+
"eval_recall_surprise": 0.0,
|
| 975 |
+
"eval_runtime": 53.1901,
|
| 976 |
+
"eval_samples_per_second": 449.407,
|
| 977 |
+
"eval_steps_per_second": 14.044,
|
| 978 |
+
"step": 31380
|
| 979 |
+
},
|
| 980 |
+
{
|
| 981 |
+
"epoch": 18.78,
|
| 982 |
+
"learning_rate": 0.00026247703420330707,
|
| 983 |
+
"loss": 1.4962,
|
| 984 |
+
"step": 31425
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"epoch": 19.03,
|
| 988 |
+
"learning_rate": 0.0002519316437218433,
|
| 989 |
+
"loss": 1.4966,
|
| 990 |
+
"step": 31844
|
| 991 |
+
},
|
| 992 |
+
{
|
| 993 |
+
"epoch": 19.28,
|
| 994 |
+
"learning_rate": 0.00024138625324037955,
|
| 995 |
+
"loss": 1.4861,
|
| 996 |
+
"step": 32263
|
| 997 |
+
},
|
| 998 |
+
{
|
| 999 |
+
"epoch": 19.53,
|
| 1000 |
+
"learning_rate": 0.00023084086275891577,
|
| 1001 |
+
"loss": 1.4909,
|
| 1002 |
+
"step": 32682
|
| 1003 |
+
},
|
| 1004 |
+
{
|
| 1005 |
+
"epoch": 19.79,
|
| 1006 |
+
"learning_rate": 0.000220295472277452,
|
| 1007 |
+
"loss": 1.4895,
|
| 1008 |
+
"step": 33101
|
| 1009 |
+
},
|
| 1010 |
+
{
|
| 1011 |
+
"epoch": 20.01,
|
| 1012 |
+
"eval_accuracy": 0.4471636546184739,
|
| 1013 |
+
"eval_b_acc": 0.14285714285714285,
|
| 1014 |
+
"eval_f1": 0.2763410114310333,
|
| 1015 |
+
"eval_f1_anger": 0.0,
|
| 1016 |
+
"eval_f1_disgust": 0.0,
|
| 1017 |
+
"eval_f1_fear": 0.0,
|
| 1018 |
+
"eval_f1_joy": 0.0,
|
| 1019 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 1020 |
+
"eval_f1_sadness": 0.0,
|
| 1021 |
+
"eval_f1_surprise": 0.0,
|
| 1022 |
+
"eval_loss": 1.5502026081085205,
|
| 1023 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 1024 |
+
"eval_prec": 0.1999553340117498,
|
| 1025 |
+
"eval_prec_anger": 0.0,
|
| 1026 |
+
"eval_prec_disgust": 0.0,
|
| 1027 |
+
"eval_prec_fear": 0.0,
|
| 1028 |
+
"eval_prec_joy": 0.0,
|
| 1029 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 1030 |
+
"eval_prec_sadness": 0.0,
|
| 1031 |
+
"eval_prec_surprise": 0.0,
|
| 1032 |
+
"eval_recall": 0.4471636546184739,
|
| 1033 |
+
"eval_recall_anger": 0.0,
|
| 1034 |
+
"eval_recall_disgust": 0.0,
|
| 1035 |
+
"eval_recall_fear": 0.0,
|
| 1036 |
+
"eval_recall_joy": 0.0,
|
| 1037 |
+
"eval_recall_neutral": 1.0,
|
| 1038 |
+
"eval_recall_sadness": 0.0,
|
| 1039 |
+
"eval_recall_surprise": 0.0,
|
| 1040 |
+
"eval_runtime": 53.1794,
|
| 1041 |
+
"eval_samples_per_second": 449.498,
|
| 1042 |
+
"eval_steps_per_second": 14.047,
|
| 1043 |
+
"step": 33472
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 20.04,
|
| 1047 |
+
"learning_rate": 0.00020975008179598823,
|
| 1048 |
+
"loss": 1.5021,
|
| 1049 |
+
"step": 33520
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 20.29,
|
| 1053 |
+
"learning_rate": 0.00019920469131452445,
|
| 1054 |
+
"loss": 1.491,
|
| 1055 |
+
"step": 33939
|
| 1056 |
+
},
|
| 1057 |
+
{
|
| 1058 |
+
"epoch": 20.54,
|
| 1059 |
+
"learning_rate": 0.00018865930083306068,
|
| 1060 |
+
"loss": 1.4844,
|
| 1061 |
+
"step": 34358
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 20.79,
|
| 1065 |
+
"learning_rate": 0.0001781139103515969,
|
| 1066 |
+
"loss": 1.4874,
|
| 1067 |
+
"step": 34777
|
| 1068 |
+
},
|
| 1069 |
+
{
|
| 1070 |
+
"epoch": 21.04,
|
| 1071 |
+
"learning_rate": 0.00016756851987013313,
|
| 1072 |
+
"loss": 1.4946,
|
| 1073 |
+
"step": 35196
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"epoch": 21.26,
|
| 1077 |
+
"eval_accuracy": 0.4471636546184739,
|
| 1078 |
+
"eval_b_acc": 0.14285714285714285,
|
| 1079 |
+
"eval_f1": 0.2763410114310333,
|
| 1080 |
+
"eval_f1_anger": 0.0,
|
| 1081 |
+
"eval_f1_disgust": 0.0,
|
| 1082 |
+
"eval_f1_fear": 0.0,
|
| 1083 |
+
"eval_f1_joy": 0.0,
|
| 1084 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 1085 |
+
"eval_f1_sadness": 0.0,
|
| 1086 |
+
"eval_f1_surprise": 0.0,
|
| 1087 |
+
"eval_loss": 1.5398402214050293,
|
| 1088 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 1089 |
+
"eval_prec": 0.1999553340117498,
|
| 1090 |
+
"eval_prec_anger": 0.0,
|
| 1091 |
+
"eval_prec_disgust": 0.0,
|
| 1092 |
+
"eval_prec_fear": 0.0,
|
| 1093 |
+
"eval_prec_joy": 0.0,
|
| 1094 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 1095 |
+
"eval_prec_sadness": 0.0,
|
| 1096 |
+
"eval_prec_surprise": 0.0,
|
| 1097 |
+
"eval_recall": 0.4471636546184739,
|
| 1098 |
+
"eval_recall_anger": 0.0,
|
| 1099 |
+
"eval_recall_disgust": 0.0,
|
| 1100 |
+
"eval_recall_fear": 0.0,
|
| 1101 |
+
"eval_recall_joy": 0.0,
|
| 1102 |
+
"eval_recall_neutral": 1.0,
|
| 1103 |
+
"eval_recall_sadness": 0.0,
|
| 1104 |
+
"eval_recall_surprise": 0.0,
|
| 1105 |
+
"eval_runtime": 53.2578,
|
| 1106 |
+
"eval_samples_per_second": 448.836,
|
| 1107 |
+
"eval_steps_per_second": 14.026,
|
| 1108 |
+
"step": 35564
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"epoch": 21.29,
|
| 1112 |
+
"learning_rate": 0.00015702312938866935,
|
| 1113 |
+
"loss": 1.4867,
|
| 1114 |
+
"step": 35615
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"epoch": 21.54,
|
| 1118 |
+
"learning_rate": 0.00014647773890720558,
|
| 1119 |
+
"loss": 1.4851,
|
| 1120 |
+
"step": 36034
|
| 1121 |
+
},
|
| 1122 |
+
{
|
| 1123 |
+
"epoch": 21.79,
|
| 1124 |
+
"learning_rate": 0.00013595751642211766,
|
| 1125 |
+
"loss": 1.4809,
|
| 1126 |
+
"step": 36453
|
| 1127 |
+
},
|
| 1128 |
+
{
|
| 1129 |
+
"epoch": 22.04,
|
| 1130 |
+
"learning_rate": 0.00012541212594065389,
|
| 1131 |
+
"loss": 1.4865,
|
| 1132 |
+
"step": 36872
|
| 1133 |
+
},
|
| 1134 |
+
{
|
| 1135 |
+
"epoch": 22.29,
|
| 1136 |
+
"learning_rate": 0.0001148667354591901,
|
| 1137 |
+
"loss": 1.4754,
|
| 1138 |
+
"step": 37291
|
| 1139 |
+
},
|
| 1140 |
+
{
|
| 1141 |
+
"epoch": 22.51,
|
| 1142 |
+
"eval_accuracy": 0.4471636546184739,
|
| 1143 |
+
"eval_b_acc": 0.14285714285714285,
|
| 1144 |
+
"eval_f1": 0.2763410114310333,
|
| 1145 |
+
"eval_f1_anger": 0.0,
|
| 1146 |
+
"eval_f1_disgust": 0.0,
|
| 1147 |
+
"eval_f1_fear": 0.0,
|
| 1148 |
+
"eval_f1_joy": 0.0,
|
| 1149 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 1150 |
+
"eval_f1_sadness": 0.0,
|
| 1151 |
+
"eval_f1_surprise": 0.0,
|
| 1152 |
+
"eval_loss": 1.5307480096817017,
|
| 1153 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 1154 |
+
"eval_prec": 0.1999553340117498,
|
| 1155 |
+
"eval_prec_anger": 0.0,
|
| 1156 |
+
"eval_prec_disgust": 0.0,
|
| 1157 |
+
"eval_prec_fear": 0.0,
|
| 1158 |
+
"eval_prec_joy": 0.0,
|
| 1159 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 1160 |
+
"eval_prec_sadness": 0.0,
|
| 1161 |
+
"eval_prec_surprise": 0.0,
|
| 1162 |
+
"eval_recall": 0.4471636546184739,
|
| 1163 |
+
"eval_recall_anger": 0.0,
|
| 1164 |
+
"eval_recall_disgust": 0.0,
|
| 1165 |
+
"eval_recall_fear": 0.0,
|
| 1166 |
+
"eval_recall_joy": 0.0,
|
| 1167 |
+
"eval_recall_neutral": 1.0,
|
| 1168 |
+
"eval_recall_sadness": 0.0,
|
| 1169 |
+
"eval_recall_surprise": 0.0,
|
| 1170 |
+
"eval_runtime": 53.2695,
|
| 1171 |
+
"eval_samples_per_second": 448.737,
|
| 1172 |
+
"eval_steps_per_second": 14.023,
|
| 1173 |
+
"step": 37656
|
| 1174 |
+
},
|
| 1175 |
+
{
|
| 1176 |
+
"epoch": 22.54,
|
| 1177 |
+
"learning_rate": 0.00010432134497772632,
|
| 1178 |
+
"loss": 1.4855,
|
| 1179 |
+
"step": 37710
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"epoch": 22.79,
|
| 1183 |
+
"learning_rate": 9.380112249263837e-05,
|
| 1184 |
+
"loss": 1.4831,
|
| 1185 |
+
"step": 38129
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 23.04,
|
| 1189 |
+
"learning_rate": 8.325573201117459e-05,
|
| 1190 |
+
"loss": 1.4884,
|
| 1191 |
+
"step": 38548
|
| 1192 |
+
},
|
| 1193 |
+
{
|
| 1194 |
+
"epoch": 23.29,
|
| 1195 |
+
"learning_rate": 7.271034152971082e-05,
|
| 1196 |
+
"loss": 1.4794,
|
| 1197 |
+
"step": 38967
|
| 1198 |
+
},
|
| 1199 |
+
{
|
| 1200 |
+
"epoch": 23.54,
|
| 1201 |
+
"learning_rate": 6.216495104824706e-05,
|
| 1202 |
+
"loss": 1.4824,
|
| 1203 |
+
"step": 39386
|
| 1204 |
+
},
|
| 1205 |
+
{
|
| 1206 |
+
"epoch": 23.76,
|
| 1207 |
+
"eval_accuracy": 0.4471636546184739,
|
| 1208 |
+
"eval_b_acc": 0.14285714285714285,
|
| 1209 |
+
"eval_f1": 0.2763410114310333,
|
| 1210 |
+
"eval_f1_anger": 0.0,
|
| 1211 |
+
"eval_f1_disgust": 0.0,
|
| 1212 |
+
"eval_f1_fear": 0.0,
|
| 1213 |
+
"eval_f1_joy": 0.0,
|
| 1214 |
+
"eval_f1_neutral": 0.6179862978059145,
|
| 1215 |
+
"eval_f1_sadness": 0.0,
|
| 1216 |
+
"eval_f1_surprise": 0.0,
|
| 1217 |
+
"eval_loss": 1.535596489906311,
|
| 1218 |
+
"eval_micro_f1": 0.4471636546184739,
|
| 1219 |
+
"eval_prec": 0.1999553340117498,
|
| 1220 |
+
"eval_prec_anger": 0.0,
|
| 1221 |
+
"eval_prec_disgust": 0.0,
|
| 1222 |
+
"eval_prec_fear": 0.0,
|
| 1223 |
+
"eval_prec_joy": 0.0,
|
| 1224 |
+
"eval_prec_neutral": 0.4471636546184739,
|
| 1225 |
+
"eval_prec_sadness": 0.0,
|
| 1226 |
+
"eval_prec_surprise": 0.0,
|
| 1227 |
+
"eval_recall": 0.4471636546184739,
|
| 1228 |
+
"eval_recall_anger": 0.0,
|
| 1229 |
+
"eval_recall_disgust": 0.0,
|
| 1230 |
+
"eval_recall_fear": 0.0,
|
| 1231 |
+
"eval_recall_joy": 0.0,
|
| 1232 |
+
"eval_recall_neutral": 1.0,
|
| 1233 |
+
"eval_recall_sadness": 0.0,
|
| 1234 |
+
"eval_recall_surprise": 0.0,
|
| 1235 |
+
"eval_runtime": 52.9319,
|
| 1236 |
+
"eval_samples_per_second": 451.599,
|
| 1237 |
+
"eval_steps_per_second": 14.112,
|
| 1238 |
+
"step": 39748
|
| 1239 |
+
},
|
| 1240 |
+
{
|
| 1241 |
+
"epoch": 23.79,
|
| 1242 |
+
"learning_rate": 5.1619560566783274e-05,
|
| 1243 |
+
"loss": 1.482,
|
| 1244 |
+
"step": 39805
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"epoch": 24.04,
|
| 1248 |
+
"learning_rate": 4.1074170085319506e-05,
|
| 1249 |
+
"loss": 1.4867,
|
| 1250 |
+
"step": 40224
|
| 1251 |
+
},
|
| 1252 |
+
{
|
| 1253 |
+
"epoch": 24.29,
|
| 1254 |
+
"learning_rate": 3.052877960385574e-05,
|
| 1255 |
+
"loss": 1.4839,
|
| 1256 |
+
"step": 40643
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 24.54,
|
| 1260 |
+
"learning_rate": 1.9983389122391967e-05,
|
| 1261 |
+
"loss": 1.4788,
|
| 1262 |
+
"step": 41062
|
| 1263 |
+
},
|
| 1264 |
+
{
|
| 1265 |
+
"epoch": 24.79,
|
| 1266 |
+
"learning_rate": 9.437998640928196e-06,
|
| 1267 |
+
"loss": 1.4844,
|
| 1268 |
+
"step": 41481
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 25.0,
|
| 1272 |
+
"step": 41825,
|
| 1273 |
+
"total_flos": 1.2547148116392576e+18,
|
| 1274 |
+
"train_loss": 1.4594040881953265,
|
| 1275 |
+
"train_runtime": 28719.4719,
|
| 1276 |
+
"train_samples_per_second": 186.383,
|
| 1277 |
+
"train_steps_per_second": 1.456
|
| 1278 |
+
}
|
| 1279 |
+
],
|
| 1280 |
+
"logging_steps": 419,
|
| 1281 |
+
"max_steps": 41825,
|
| 1282 |
+
"num_train_epochs": 25,
|
| 1283 |
+
"save_steps": 4183,
|
| 1284 |
+
"total_flos": 1.2547148116392576e+18,
|
| 1285 |
+
"trial_name": null,
|
| 1286 |
+
"trial_params": null
|
| 1287 |
+
}
|