End of training
Browse files- README.md +12 -12
- all_results.json +22 -0
- eval_results.json +17 -0
- runs/Oct17_11-15-19_tech/events.out.tfevents.1760719385.tech.80033.1 +3 -0
- train_results.json +8 -0
- trainer_state.json +1153 -0
README.md
CHANGED
|
@@ -23,7 +23,7 @@ model-index:
|
|
| 23 |
metrics:
|
| 24 |
- name: Accuracy
|
| 25 |
type: accuracy
|
| 26 |
-
value: 0.
|
| 27 |
---
|
| 28 |
|
| 29 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -33,17 +33,17 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 33 |
|
| 34 |
This model is a fine-tuned version of [google/vit-base-patch16-384](https://huggingface.co/google/vit-base-patch16-384) on the imagefolder dataset.
|
| 35 |
It achieves the following results on the evaluation set:
|
| 36 |
-
- Loss: 0.
|
| 37 |
-
- Accuracy: 0.
|
| 38 |
-
- F1 Macro: 0.
|
| 39 |
-
- Precision Macro: 0.
|
| 40 |
-
- Recall Macro: 0.
|
| 41 |
-
- Precision Dry: 0.
|
| 42 |
-
- Recall Dry: 0.
|
| 43 |
-
- F1 Dry: 0.
|
| 44 |
-
- Precision Humid: 0.
|
| 45 |
-
- Recall Humid: 0.
|
| 46 |
-
- F1 Humid: 0.
|
| 47 |
|
| 48 |
## Model description
|
| 49 |
|
|
|
|
| 23 |
metrics:
|
| 24 |
- name: Accuracy
|
| 25 |
type: accuracy
|
| 26 |
+
value: 0.9411764705882353
|
| 27 |
---
|
| 28 |
|
| 29 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 33 |
|
| 34 |
This model is a fine-tuned version of [google/vit-base-patch16-384](https://huggingface.co/google/vit-base-patch16-384) on the imagefolder dataset.
|
| 35 |
It achieves the following results on the evaluation set:
|
| 36 |
+
- Loss: 0.4611
|
| 37 |
+
- Accuracy: 0.9412
|
| 38 |
+
- F1 Macro: 0.9399
|
| 39 |
+
- Precision Macro: 0.9399
|
| 40 |
+
- Recall Macro: 0.9399
|
| 41 |
+
- Precision Dry: 0.9310
|
| 42 |
+
- Recall Dry: 0.9310
|
| 43 |
+
- F1 Dry: 0.9310
|
| 44 |
+
- Precision Humid: 0.9487
|
| 45 |
+
- Recall Humid: 0.9487
|
| 46 |
+
- F1 Humid: 0.9487
|
| 47 |
|
| 48 |
## Model description
|
| 49 |
|
all_results.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 50.0,
|
| 3 |
+
"eval_accuracy": 0.9411764705882353,
|
| 4 |
+
"eval_f1_dry": 0.9310344827586207,
|
| 5 |
+
"eval_f1_humid": 0.9487179487179487,
|
| 6 |
+
"eval_f1_macro": 0.9398762157382847,
|
| 7 |
+
"eval_loss": 0.4611157774925232,
|
| 8 |
+
"eval_precision_dry": 0.9310344827586207,
|
| 9 |
+
"eval_precision_humid": 0.9487179487179487,
|
| 10 |
+
"eval_precision_macro": 0.9398762157382847,
|
| 11 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 12 |
+
"eval_recall_humid": 0.9487179487179487,
|
| 13 |
+
"eval_recall_macro": 0.9398762157382847,
|
| 14 |
+
"eval_runtime": 2.2225,
|
| 15 |
+
"eval_samples_per_second": 30.596,
|
| 16 |
+
"eval_steps_per_second": 2.25,
|
| 17 |
+
"total_flos": 4.3301918339334144e+18,
|
| 18 |
+
"train_loss": 0.04831398472228708,
|
| 19 |
+
"train_runtime": 1645.2014,
|
| 20 |
+
"train_samples_per_second": 11.518,
|
| 21 |
+
"train_steps_per_second": 0.182
|
| 22 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 50.0,
|
| 3 |
+
"eval_accuracy": 0.9411764705882353,
|
| 4 |
+
"eval_f1_dry": 0.9310344827586207,
|
| 5 |
+
"eval_f1_humid": 0.9487179487179487,
|
| 6 |
+
"eval_f1_macro": 0.9398762157382847,
|
| 7 |
+
"eval_loss": 0.4611157774925232,
|
| 8 |
+
"eval_precision_dry": 0.9310344827586207,
|
| 9 |
+
"eval_precision_humid": 0.9487179487179487,
|
| 10 |
+
"eval_precision_macro": 0.9398762157382847,
|
| 11 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 12 |
+
"eval_recall_humid": 0.9487179487179487,
|
| 13 |
+
"eval_recall_macro": 0.9398762157382847,
|
| 14 |
+
"eval_runtime": 2.2225,
|
| 15 |
+
"eval_samples_per_second": 30.596,
|
| 16 |
+
"eval_steps_per_second": 2.25
|
| 17 |
+
}
|
runs/Oct17_11-15-19_tech/events.out.tfevents.1760719385.tech.80033.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a18ee7148cbe3e463462a53856e27b7a6864881240bffe63e1abe4bbe90917f
|
| 3 |
+
size 906
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 50.0,
|
| 3 |
+
"total_flos": 4.3301918339334144e+18,
|
| 4 |
+
"train_loss": 0.04831398472228708,
|
| 5 |
+
"train_runtime": 1645.2014,
|
| 6 |
+
"train_samples_per_second": 11.518,
|
| 7 |
+
"train_steps_per_second": 0.182
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1153 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_global_step": 96,
|
| 3 |
+
"best_metric": 0.9411764705882353,
|
| 4 |
+
"best_model_checkpoint": "vit-base-patch16-384-finetuned-humid-binary-2/checkpoint-96",
|
| 5 |
+
"epoch": 50.0,
|
| 6 |
+
"eval_steps": 500,
|
| 7 |
+
"global_step": 300,
|
| 8 |
+
"is_hyper_param_search": false,
|
| 9 |
+
"is_local_process_zero": true,
|
| 10 |
+
"is_world_process_zero": true,
|
| 11 |
+
"log_history": [
|
| 12 |
+
{
|
| 13 |
+
"epoch": 1.0,
|
| 14 |
+
"eval_accuracy": 0.7205882352941176,
|
| 15 |
+
"eval_f1_dry": 0.6885245901639344,
|
| 16 |
+
"eval_f1_humid": 0.7466666666666667,
|
| 17 |
+
"eval_f1_macro": 0.7175956284153006,
|
| 18 |
+
"eval_loss": 0.614932119846344,
|
| 19 |
+
"eval_precision_dry": 0.65625,
|
| 20 |
+
"eval_precision_humid": 0.7777777777777778,
|
| 21 |
+
"eval_precision_macro": 0.7170138888888888,
|
| 22 |
+
"eval_recall_dry": 0.7241379310344828,
|
| 23 |
+
"eval_recall_humid": 0.717948717948718,
|
| 24 |
+
"eval_recall_macro": 0.7210433244916004,
|
| 25 |
+
"eval_runtime": 2.2698,
|
| 26 |
+
"eval_samples_per_second": 29.959,
|
| 27 |
+
"eval_steps_per_second": 2.203,
|
| 28 |
+
"step": 6
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"epoch": 1.6666666666666665,
|
| 32 |
+
"grad_norm": 5.209638595581055,
|
| 33 |
+
"learning_rate": 1.5e-05,
|
| 34 |
+
"loss": 0.6101,
|
| 35 |
+
"step": 10
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"epoch": 2.0,
|
| 39 |
+
"eval_accuracy": 0.8823529411764706,
|
| 40 |
+
"eval_f1_dry": 0.8620689655172413,
|
| 41 |
+
"eval_f1_humid": 0.8974358974358975,
|
| 42 |
+
"eval_f1_macro": 0.8797524314765695,
|
| 43 |
+
"eval_loss": 0.3864850401878357,
|
| 44 |
+
"eval_precision_dry": 0.8620689655172413,
|
| 45 |
+
"eval_precision_humid": 0.8974358974358975,
|
| 46 |
+
"eval_precision_macro": 0.8797524314765695,
|
| 47 |
+
"eval_recall_dry": 0.8620689655172413,
|
| 48 |
+
"eval_recall_humid": 0.8974358974358975,
|
| 49 |
+
"eval_recall_macro": 0.8797524314765695,
|
| 50 |
+
"eval_runtime": 2.2426,
|
| 51 |
+
"eval_samples_per_second": 30.322,
|
| 52 |
+
"eval_steps_per_second": 2.23,
|
| 53 |
+
"step": 12
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"epoch": 3.0,
|
| 57 |
+
"eval_accuracy": 0.8382352941176471,
|
| 58 |
+
"eval_f1_dry": 0.8307692307692308,
|
| 59 |
+
"eval_f1_humid": 0.8450704225352113,
|
| 60 |
+
"eval_f1_macro": 0.837919826652221,
|
| 61 |
+
"eval_loss": 0.35972756147384644,
|
| 62 |
+
"eval_precision_dry": 0.75,
|
| 63 |
+
"eval_precision_humid": 0.9375,
|
| 64 |
+
"eval_precision_macro": 0.84375,
|
| 65 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 66 |
+
"eval_recall_humid": 0.7692307692307693,
|
| 67 |
+
"eval_recall_macro": 0.850132625994695,
|
| 68 |
+
"eval_runtime": 2.2171,
|
| 69 |
+
"eval_samples_per_second": 30.671,
|
| 70 |
+
"eval_steps_per_second": 2.255,
|
| 71 |
+
"step": 18
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 3.3333333333333335,
|
| 75 |
+
"grad_norm": 3.4679043292999268,
|
| 76 |
+
"learning_rate": 3.1666666666666666e-05,
|
| 77 |
+
"loss": 0.2613,
|
| 78 |
+
"step": 20
|
| 79 |
+
},
|
| 80 |
+
{
|
| 81 |
+
"epoch": 4.0,
|
| 82 |
+
"eval_accuracy": 0.8235294117647058,
|
| 83 |
+
"eval_f1_dry": 0.8235294117647058,
|
| 84 |
+
"eval_f1_humid": 0.8235294117647058,
|
| 85 |
+
"eval_f1_macro": 0.8235294117647058,
|
| 86 |
+
"eval_loss": 0.5013691186904907,
|
| 87 |
+
"eval_precision_dry": 0.717948717948718,
|
| 88 |
+
"eval_precision_humid": 0.9655172413793104,
|
| 89 |
+
"eval_precision_macro": 0.8417329796640142,
|
| 90 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 91 |
+
"eval_recall_humid": 0.717948717948718,
|
| 92 |
+
"eval_recall_macro": 0.8417329796640142,
|
| 93 |
+
"eval_runtime": 2.2748,
|
| 94 |
+
"eval_samples_per_second": 29.893,
|
| 95 |
+
"eval_steps_per_second": 2.198,
|
| 96 |
+
"step": 24
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"epoch": 5.0,
|
| 100 |
+
"grad_norm": 4.22995662689209,
|
| 101 |
+
"learning_rate": 4.8333333333333334e-05,
|
| 102 |
+
"loss": 0.1505,
|
| 103 |
+
"step": 30
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"epoch": 5.0,
|
| 107 |
+
"eval_accuracy": 0.8529411764705882,
|
| 108 |
+
"eval_f1_dry": 0.84375,
|
| 109 |
+
"eval_f1_humid": 0.8611111111111112,
|
| 110 |
+
"eval_f1_macro": 0.8524305555555556,
|
| 111 |
+
"eval_loss": 0.43828415870666504,
|
| 112 |
+
"eval_precision_dry": 0.7714285714285715,
|
| 113 |
+
"eval_precision_humid": 0.9393939393939394,
|
| 114 |
+
"eval_precision_macro": 0.8554112554112554,
|
| 115 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 116 |
+
"eval_recall_humid": 0.7948717948717948,
|
| 117 |
+
"eval_recall_macro": 0.8629531388152077,
|
| 118 |
+
"eval_runtime": 2.1723,
|
| 119 |
+
"eval_samples_per_second": 31.303,
|
| 120 |
+
"eval_steps_per_second": 2.302,
|
| 121 |
+
"step": 30
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 6.0,
|
| 125 |
+
"eval_accuracy": 0.8529411764705882,
|
| 126 |
+
"eval_f1_dry": 0.8484848484848485,
|
| 127 |
+
"eval_f1_humid": 0.8571428571428571,
|
| 128 |
+
"eval_f1_macro": 0.8528138528138528,
|
| 129 |
+
"eval_loss": 0.4248289167881012,
|
| 130 |
+
"eval_precision_dry": 0.7567567567567568,
|
| 131 |
+
"eval_precision_humid": 0.967741935483871,
|
| 132 |
+
"eval_precision_macro": 0.8622493461203139,
|
| 133 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 134 |
+
"eval_recall_humid": 0.7692307692307693,
|
| 135 |
+
"eval_recall_macro": 0.8673740053050398,
|
| 136 |
+
"eval_runtime": 2.3123,
|
| 137 |
+
"eval_samples_per_second": 29.408,
|
| 138 |
+
"eval_steps_per_second": 2.162,
|
| 139 |
+
"step": 36
|
| 140 |
+
},
|
| 141 |
+
{
|
| 142 |
+
"epoch": 6.666666666666667,
|
| 143 |
+
"grad_norm": 6.771641254425049,
|
| 144 |
+
"learning_rate": 4.8333333333333334e-05,
|
| 145 |
+
"loss": 0.1298,
|
| 146 |
+
"step": 40
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"epoch": 7.0,
|
| 150 |
+
"eval_accuracy": 0.8970588235294118,
|
| 151 |
+
"eval_f1_dry": 0.8727272727272727,
|
| 152 |
+
"eval_f1_humid": 0.9135802469135802,
|
| 153 |
+
"eval_f1_macro": 0.8931537598204264,
|
| 154 |
+
"eval_loss": 0.39324793219566345,
|
| 155 |
+
"eval_precision_dry": 0.9230769230769231,
|
| 156 |
+
"eval_precision_humid": 0.8809523809523809,
|
| 157 |
+
"eval_precision_macro": 0.902014652014652,
|
| 158 |
+
"eval_recall_dry": 0.8275862068965517,
|
| 159 |
+
"eval_recall_humid": 0.9487179487179487,
|
| 160 |
+
"eval_recall_macro": 0.8881520778072503,
|
| 161 |
+
"eval_runtime": 2.2011,
|
| 162 |
+
"eval_samples_per_second": 30.894,
|
| 163 |
+
"eval_steps_per_second": 2.272,
|
| 164 |
+
"step": 42
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 8.0,
|
| 168 |
+
"eval_accuracy": 0.8970588235294118,
|
| 169 |
+
"eval_f1_dry": 0.8852459016393442,
|
| 170 |
+
"eval_f1_humid": 0.9066666666666666,
|
| 171 |
+
"eval_f1_macro": 0.8959562841530054,
|
| 172 |
+
"eval_loss": 0.3932932913303375,
|
| 173 |
+
"eval_precision_dry": 0.84375,
|
| 174 |
+
"eval_precision_humid": 0.9444444444444444,
|
| 175 |
+
"eval_precision_macro": 0.8940972222222222,
|
| 176 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 177 |
+
"eval_recall_humid": 0.8717948717948718,
|
| 178 |
+
"eval_recall_macro": 0.9014146772767462,
|
| 179 |
+
"eval_runtime": 2.1675,
|
| 180 |
+
"eval_samples_per_second": 31.373,
|
| 181 |
+
"eval_steps_per_second": 2.307,
|
| 182 |
+
"step": 48
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 8.333333333333334,
|
| 186 |
+
"grad_norm": 2.7544643878936768,
|
| 187 |
+
"learning_rate": 4.648148148148148e-05,
|
| 188 |
+
"loss": 0.1033,
|
| 189 |
+
"step": 50
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 9.0,
|
| 193 |
+
"eval_accuracy": 0.7941176470588235,
|
| 194 |
+
"eval_f1_dry": 0.8055555555555556,
|
| 195 |
+
"eval_f1_humid": 0.78125,
|
| 196 |
+
"eval_f1_macro": 0.7934027777777778,
|
| 197 |
+
"eval_loss": 0.7160156965255737,
|
| 198 |
+
"eval_precision_dry": 0.6744186046511628,
|
| 199 |
+
"eval_precision_humid": 1.0,
|
| 200 |
+
"eval_precision_macro": 0.8372093023255813,
|
| 201 |
+
"eval_recall_dry": 1.0,
|
| 202 |
+
"eval_recall_humid": 0.6410256410256411,
|
| 203 |
+
"eval_recall_macro": 0.8205128205128205,
|
| 204 |
+
"eval_runtime": 2.1872,
|
| 205 |
+
"eval_samples_per_second": 31.09,
|
| 206 |
+
"eval_steps_per_second": 2.286,
|
| 207 |
+
"step": 54
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 10.0,
|
| 211 |
+
"grad_norm": 2.3722591400146484,
|
| 212 |
+
"learning_rate": 4.462962962962963e-05,
|
| 213 |
+
"loss": 0.1135,
|
| 214 |
+
"step": 60
|
| 215 |
+
},
|
| 216 |
+
{
|
| 217 |
+
"epoch": 10.0,
|
| 218 |
+
"eval_accuracy": 0.9264705882352942,
|
| 219 |
+
"eval_f1_dry": 0.9152542372881356,
|
| 220 |
+
"eval_f1_humid": 0.935064935064935,
|
| 221 |
+
"eval_f1_macro": 0.9251595861765354,
|
| 222 |
+
"eval_loss": 0.40363243222236633,
|
| 223 |
+
"eval_precision_dry": 0.9,
|
| 224 |
+
"eval_precision_humid": 0.9473684210526315,
|
| 225 |
+
"eval_precision_macro": 0.9236842105263158,
|
| 226 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 227 |
+
"eval_recall_humid": 0.9230769230769231,
|
| 228 |
+
"eval_recall_macro": 0.927055702917772,
|
| 229 |
+
"eval_runtime": 2.1789,
|
| 230 |
+
"eval_samples_per_second": 31.209,
|
| 231 |
+
"eval_steps_per_second": 2.295,
|
| 232 |
+
"step": 60
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 11.0,
|
| 236 |
+
"eval_accuracy": 0.9117647058823529,
|
| 237 |
+
"eval_f1_dry": 0.8928571428571429,
|
| 238 |
+
"eval_f1_humid": 0.925,
|
| 239 |
+
"eval_f1_macro": 0.9089285714285715,
|
| 240 |
+
"eval_loss": 0.4228282868862152,
|
| 241 |
+
"eval_precision_dry": 0.9259259259259259,
|
| 242 |
+
"eval_precision_humid": 0.9024390243902439,
|
| 243 |
+
"eval_precision_macro": 0.9141824751580849,
|
| 244 |
+
"eval_recall_dry": 0.8620689655172413,
|
| 245 |
+
"eval_recall_humid": 0.9487179487179487,
|
| 246 |
+
"eval_recall_macro": 0.905393457117595,
|
| 247 |
+
"eval_runtime": 2.2933,
|
| 248 |
+
"eval_samples_per_second": 29.651,
|
| 249 |
+
"eval_steps_per_second": 2.18,
|
| 250 |
+
"step": 66
|
| 251 |
+
},
|
| 252 |
+
{
|
| 253 |
+
"epoch": 11.666666666666666,
|
| 254 |
+
"grad_norm": 0.39826828241348267,
|
| 255 |
+
"learning_rate": 4.277777777777778e-05,
|
| 256 |
+
"loss": 0.0331,
|
| 257 |
+
"step": 70
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 12.0,
|
| 261 |
+
"eval_accuracy": 0.8823529411764706,
|
| 262 |
+
"eval_f1_dry": 0.875,
|
| 263 |
+
"eval_f1_humid": 0.8888888888888888,
|
| 264 |
+
"eval_f1_macro": 0.8819444444444444,
|
| 265 |
+
"eval_loss": 0.4820828139781952,
|
| 266 |
+
"eval_precision_dry": 0.8,
|
| 267 |
+
"eval_precision_humid": 0.9696969696969697,
|
| 268 |
+
"eval_precision_macro": 0.8848484848484849,
|
| 269 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 270 |
+
"eval_recall_humid": 0.8205128205128205,
|
| 271 |
+
"eval_recall_macro": 0.8930150309460654,
|
| 272 |
+
"eval_runtime": 2.2848,
|
| 273 |
+
"eval_samples_per_second": 29.762,
|
| 274 |
+
"eval_steps_per_second": 2.188,
|
| 275 |
+
"step": 72
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 13.0,
|
| 279 |
+
"eval_accuracy": 0.9117647058823529,
|
| 280 |
+
"eval_f1_dry": 0.9,
|
| 281 |
+
"eval_f1_humid": 0.9210526315789473,
|
| 282 |
+
"eval_f1_macro": 0.9105263157894736,
|
| 283 |
+
"eval_loss": 0.4474344849586487,
|
| 284 |
+
"eval_precision_dry": 0.8709677419354839,
|
| 285 |
+
"eval_precision_humid": 0.9459459459459459,
|
| 286 |
+
"eval_precision_macro": 0.9084568439407149,
|
| 287 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 288 |
+
"eval_recall_humid": 0.8974358974358975,
|
| 289 |
+
"eval_recall_macro": 0.9142351900972591,
|
| 290 |
+
"eval_runtime": 2.3596,
|
| 291 |
+
"eval_samples_per_second": 28.819,
|
| 292 |
+
"eval_steps_per_second": 2.119,
|
| 293 |
+
"step": 78
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 13.333333333333334,
|
| 297 |
+
"grad_norm": 4.4798173904418945,
|
| 298 |
+
"learning_rate": 4.092592592592593e-05,
|
| 299 |
+
"loss": 0.0142,
|
| 300 |
+
"step": 80
|
| 301 |
+
},
|
| 302 |
+
{
|
| 303 |
+
"epoch": 14.0,
|
| 304 |
+
"eval_accuracy": 0.9117647058823529,
|
| 305 |
+
"eval_f1_dry": 0.896551724137931,
|
| 306 |
+
"eval_f1_humid": 0.9230769230769231,
|
| 307 |
+
"eval_f1_macro": 0.9098143236074271,
|
| 308 |
+
"eval_loss": 0.4478212893009186,
|
| 309 |
+
"eval_precision_dry": 0.896551724137931,
|
| 310 |
+
"eval_precision_humid": 0.9230769230769231,
|
| 311 |
+
"eval_precision_macro": 0.9098143236074271,
|
| 312 |
+
"eval_recall_dry": 0.896551724137931,
|
| 313 |
+
"eval_recall_humid": 0.9230769230769231,
|
| 314 |
+
"eval_recall_macro": 0.9098143236074271,
|
| 315 |
+
"eval_runtime": 2.1905,
|
| 316 |
+
"eval_samples_per_second": 31.043,
|
| 317 |
+
"eval_steps_per_second": 2.283,
|
| 318 |
+
"step": 84
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 15.0,
|
| 322 |
+
"grad_norm": 0.2734902501106262,
|
| 323 |
+
"learning_rate": 3.9074074074074076e-05,
|
| 324 |
+
"loss": 0.0074,
|
| 325 |
+
"step": 90
|
| 326 |
+
},
|
| 327 |
+
{
|
| 328 |
+
"epoch": 15.0,
|
| 329 |
+
"eval_accuracy": 0.8823529411764706,
|
| 330 |
+
"eval_f1_dry": 0.875,
|
| 331 |
+
"eval_f1_humid": 0.8888888888888888,
|
| 332 |
+
"eval_f1_macro": 0.8819444444444444,
|
| 333 |
+
"eval_loss": 0.6301909685134888,
|
| 334 |
+
"eval_precision_dry": 0.8,
|
| 335 |
+
"eval_precision_humid": 0.9696969696969697,
|
| 336 |
+
"eval_precision_macro": 0.8848484848484849,
|
| 337 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 338 |
+
"eval_recall_humid": 0.8205128205128205,
|
| 339 |
+
"eval_recall_macro": 0.8930150309460654,
|
| 340 |
+
"eval_runtime": 2.2,
|
| 341 |
+
"eval_samples_per_second": 30.909,
|
| 342 |
+
"eval_steps_per_second": 2.273,
|
| 343 |
+
"step": 90
|
| 344 |
+
},
|
| 345 |
+
{
|
| 346 |
+
"epoch": 16.0,
|
| 347 |
+
"eval_accuracy": 0.9411764705882353,
|
| 348 |
+
"eval_f1_dry": 0.9310344827586207,
|
| 349 |
+
"eval_f1_humid": 0.9487179487179487,
|
| 350 |
+
"eval_f1_macro": 0.9398762157382847,
|
| 351 |
+
"eval_loss": 0.4611157774925232,
|
| 352 |
+
"eval_precision_dry": 0.9310344827586207,
|
| 353 |
+
"eval_precision_humid": 0.9487179487179487,
|
| 354 |
+
"eval_precision_macro": 0.9398762157382847,
|
| 355 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 356 |
+
"eval_recall_humid": 0.9487179487179487,
|
| 357 |
+
"eval_recall_macro": 0.9398762157382847,
|
| 358 |
+
"eval_runtime": 2.2029,
|
| 359 |
+
"eval_samples_per_second": 30.868,
|
| 360 |
+
"eval_steps_per_second": 2.27,
|
| 361 |
+
"step": 96
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 16.666666666666668,
|
| 365 |
+
"grad_norm": 0.033265743404626846,
|
| 366 |
+
"learning_rate": 3.722222222222222e-05,
|
| 367 |
+
"loss": 0.0114,
|
| 368 |
+
"step": 100
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 17.0,
|
| 372 |
+
"eval_accuracy": 0.8823529411764706,
|
| 373 |
+
"eval_f1_dry": 0.875,
|
| 374 |
+
"eval_f1_humid": 0.8888888888888888,
|
| 375 |
+
"eval_f1_macro": 0.8819444444444444,
|
| 376 |
+
"eval_loss": 0.6472479104995728,
|
| 377 |
+
"eval_precision_dry": 0.8,
|
| 378 |
+
"eval_precision_humid": 0.9696969696969697,
|
| 379 |
+
"eval_precision_macro": 0.8848484848484849,
|
| 380 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 381 |
+
"eval_recall_humid": 0.8205128205128205,
|
| 382 |
+
"eval_recall_macro": 0.8930150309460654,
|
| 383 |
+
"eval_runtime": 2.1737,
|
| 384 |
+
"eval_samples_per_second": 31.282,
|
| 385 |
+
"eval_steps_per_second": 2.3,
|
| 386 |
+
"step": 102
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"epoch": 18.0,
|
| 390 |
+
"eval_accuracy": 0.9117647058823529,
|
| 391 |
+
"eval_f1_dry": 0.8928571428571429,
|
| 392 |
+
"eval_f1_humid": 0.925,
|
| 393 |
+
"eval_f1_macro": 0.9089285714285715,
|
| 394 |
+
"eval_loss": 0.4689759910106659,
|
| 395 |
+
"eval_precision_dry": 0.9259259259259259,
|
| 396 |
+
"eval_precision_humid": 0.9024390243902439,
|
| 397 |
+
"eval_precision_macro": 0.9141824751580849,
|
| 398 |
+
"eval_recall_dry": 0.8620689655172413,
|
| 399 |
+
"eval_recall_humid": 0.9487179487179487,
|
| 400 |
+
"eval_recall_macro": 0.905393457117595,
|
| 401 |
+
"eval_runtime": 2.1917,
|
| 402 |
+
"eval_samples_per_second": 31.027,
|
| 403 |
+
"eval_steps_per_second": 2.281,
|
| 404 |
+
"step": 108
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 18.333333333333332,
|
| 408 |
+
"grad_norm": 0.9351687431335449,
|
| 409 |
+
"learning_rate": 3.537037037037037e-05,
|
| 410 |
+
"loss": 0.0054,
|
| 411 |
+
"step": 110
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 19.0,
|
| 415 |
+
"eval_accuracy": 0.8676470588235294,
|
| 416 |
+
"eval_f1_dry": 0.8615384615384616,
|
| 417 |
+
"eval_f1_humid": 0.8732394366197183,
|
| 418 |
+
"eval_f1_macro": 0.8673889490790899,
|
| 419 |
+
"eval_loss": 0.7610320448875427,
|
| 420 |
+
"eval_precision_dry": 0.7777777777777778,
|
| 421 |
+
"eval_precision_humid": 0.96875,
|
| 422 |
+
"eval_precision_macro": 0.8732638888888888,
|
| 423 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 424 |
+
"eval_recall_humid": 0.7948717948717948,
|
| 425 |
+
"eval_recall_macro": 0.8801945181255526,
|
| 426 |
+
"eval_runtime": 2.1632,
|
| 427 |
+
"eval_samples_per_second": 31.435,
|
| 428 |
+
"eval_steps_per_second": 2.311,
|
| 429 |
+
"step": 114
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 20.0,
|
| 433 |
+
"grad_norm": 0.03664712235331535,
|
| 434 |
+
"learning_rate": 3.351851851851852e-05,
|
| 435 |
+
"loss": 0.0024,
|
| 436 |
+
"step": 120
|
| 437 |
+
},
|
| 438 |
+
{
|
| 439 |
+
"epoch": 20.0,
|
| 440 |
+
"eval_accuracy": 0.9117647058823529,
|
| 441 |
+
"eval_f1_dry": 0.9,
|
| 442 |
+
"eval_f1_humid": 0.9210526315789473,
|
| 443 |
+
"eval_f1_macro": 0.9105263157894736,
|
| 444 |
+
"eval_loss": 0.5395424365997314,
|
| 445 |
+
"eval_precision_dry": 0.8709677419354839,
|
| 446 |
+
"eval_precision_humid": 0.9459459459459459,
|
| 447 |
+
"eval_precision_macro": 0.9084568439407149,
|
| 448 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 449 |
+
"eval_recall_humid": 0.8974358974358975,
|
| 450 |
+
"eval_recall_macro": 0.9142351900972591,
|
| 451 |
+
"eval_runtime": 2.1855,
|
| 452 |
+
"eval_samples_per_second": 31.114,
|
| 453 |
+
"eval_steps_per_second": 2.288,
|
| 454 |
+
"step": 120
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"epoch": 21.0,
|
| 458 |
+
"eval_accuracy": 0.8970588235294118,
|
| 459 |
+
"eval_f1_dry": 0.8852459016393442,
|
| 460 |
+
"eval_f1_humid": 0.9066666666666666,
|
| 461 |
+
"eval_f1_macro": 0.8959562841530054,
|
| 462 |
+
"eval_loss": 0.6274784207344055,
|
| 463 |
+
"eval_precision_dry": 0.84375,
|
| 464 |
+
"eval_precision_humid": 0.9444444444444444,
|
| 465 |
+
"eval_precision_macro": 0.8940972222222222,
|
| 466 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 467 |
+
"eval_recall_humid": 0.8717948717948718,
|
| 468 |
+
"eval_recall_macro": 0.9014146772767462,
|
| 469 |
+
"eval_runtime": 2.2125,
|
| 470 |
+
"eval_samples_per_second": 30.734,
|
| 471 |
+
"eval_steps_per_second": 2.26,
|
| 472 |
+
"step": 126
|
| 473 |
+
},
|
| 474 |
+
{
|
| 475 |
+
"epoch": 21.666666666666668,
|
| 476 |
+
"grad_norm": 0.053658369928598404,
|
| 477 |
+
"learning_rate": 3.1666666666666666e-05,
|
| 478 |
+
"loss": 0.0024,
|
| 479 |
+
"step": 130
|
| 480 |
+
},
|
| 481 |
+
{
|
| 482 |
+
"epoch": 22.0,
|
| 483 |
+
"eval_accuracy": 0.8676470588235294,
|
| 484 |
+
"eval_f1_dry": 0.8615384615384616,
|
| 485 |
+
"eval_f1_humid": 0.8732394366197183,
|
| 486 |
+
"eval_f1_macro": 0.8673889490790899,
|
| 487 |
+
"eval_loss": 0.7972272634506226,
|
| 488 |
+
"eval_precision_dry": 0.7777777777777778,
|
| 489 |
+
"eval_precision_humid": 0.96875,
|
| 490 |
+
"eval_precision_macro": 0.8732638888888888,
|
| 491 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 492 |
+
"eval_recall_humid": 0.7948717948717948,
|
| 493 |
+
"eval_recall_macro": 0.8801945181255526,
|
| 494 |
+
"eval_runtime": 2.2057,
|
| 495 |
+
"eval_samples_per_second": 30.829,
|
| 496 |
+
"eval_steps_per_second": 2.267,
|
| 497 |
+
"step": 132
|
| 498 |
+
},
|
| 499 |
+
{
|
| 500 |
+
"epoch": 23.0,
|
| 501 |
+
"eval_accuracy": 0.9117647058823529,
|
| 502 |
+
"eval_f1_dry": 0.9,
|
| 503 |
+
"eval_f1_humid": 0.9210526315789473,
|
| 504 |
+
"eval_f1_macro": 0.9105263157894736,
|
| 505 |
+
"eval_loss": 0.5465890169143677,
|
| 506 |
+
"eval_precision_dry": 0.8709677419354839,
|
| 507 |
+
"eval_precision_humid": 0.9459459459459459,
|
| 508 |
+
"eval_precision_macro": 0.9084568439407149,
|
| 509 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 510 |
+
"eval_recall_humid": 0.8974358974358975,
|
| 511 |
+
"eval_recall_macro": 0.9142351900972591,
|
| 512 |
+
"eval_runtime": 2.2093,
|
| 513 |
+
"eval_samples_per_second": 30.779,
|
| 514 |
+
"eval_steps_per_second": 2.263,
|
| 515 |
+
"step": 138
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 23.333333333333332,
|
| 519 |
+
"grad_norm": 0.01843937672674656,
|
| 520 |
+
"learning_rate": 2.981481481481482e-05,
|
| 521 |
+
"loss": 0.0009,
|
| 522 |
+
"step": 140
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 24.0,
|
| 526 |
+
"eval_accuracy": 0.8970588235294118,
|
| 527 |
+
"eval_f1_dry": 0.8852459016393442,
|
| 528 |
+
"eval_f1_humid": 0.9066666666666666,
|
| 529 |
+
"eval_f1_macro": 0.8959562841530054,
|
| 530 |
+
"eval_loss": 0.5303817987442017,
|
| 531 |
+
"eval_precision_dry": 0.84375,
|
| 532 |
+
"eval_precision_humid": 0.9444444444444444,
|
| 533 |
+
"eval_precision_macro": 0.8940972222222222,
|
| 534 |
+
"eval_recall_dry": 0.9310344827586207,
|
| 535 |
+
"eval_recall_humid": 0.8717948717948718,
|
| 536 |
+
"eval_recall_macro": 0.9014146772767462,
|
| 537 |
+
"eval_runtime": 2.1789,
|
| 538 |
+
"eval_samples_per_second": 31.208,
|
| 539 |
+
"eval_steps_per_second": 2.295,
|
| 540 |
+
"step": 144
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"epoch": 25.0,
|
| 544 |
+
"grad_norm": 0.0267292782664299,
|
| 545 |
+
"learning_rate": 2.7962962962962965e-05,
|
| 546 |
+
"loss": 0.0012,
|
| 547 |
+
"step": 150
|
| 548 |
+
},
|
| 549 |
+
{
|
| 550 |
+
"epoch": 25.0,
|
| 551 |
+
"eval_accuracy": 0.8970588235294118,
|
| 552 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 553 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 554 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 555 |
+
"eval_loss": 0.656761646270752,
|
| 556 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 557 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 558 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 559 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 560 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 561 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 562 |
+
"eval_runtime": 2.288,
|
| 563 |
+
"eval_samples_per_second": 29.721,
|
| 564 |
+
"eval_steps_per_second": 2.185,
|
| 565 |
+
"step": 150
|
| 566 |
+
},
|
| 567 |
+
{
|
| 568 |
+
"epoch": 26.0,
|
| 569 |
+
"eval_accuracy": 0.8970588235294118,
|
| 570 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 571 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 572 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 573 |
+
"eval_loss": 0.613365650177002,
|
| 574 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 575 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 576 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 577 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 578 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 579 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 580 |
+
"eval_runtime": 2.3183,
|
| 581 |
+
"eval_samples_per_second": 29.332,
|
| 582 |
+
"eval_steps_per_second": 2.157,
|
| 583 |
+
"step": 156
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"epoch": 26.666666666666668,
|
| 587 |
+
"grad_norm": 0.007465701084583998,
|
| 588 |
+
"learning_rate": 2.6111111111111114e-05,
|
| 589 |
+
"loss": 0.0004,
|
| 590 |
+
"step": 160
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 27.0,
|
| 594 |
+
"eval_accuracy": 0.8970588235294118,
|
| 595 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 596 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 597 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 598 |
+
"eval_loss": 0.6093538999557495,
|
| 599 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 600 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 601 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 602 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 603 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 604 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 605 |
+
"eval_runtime": 2.2556,
|
| 606 |
+
"eval_samples_per_second": 30.148,
|
| 607 |
+
"eval_steps_per_second": 2.217,
|
| 608 |
+
"step": 162
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 28.0,
|
| 612 |
+
"eval_accuracy": 0.8970588235294118,
|
| 613 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 614 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 615 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 616 |
+
"eval_loss": 0.615250825881958,
|
| 617 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 618 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 619 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 620 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 621 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 622 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 623 |
+
"eval_runtime": 2.1738,
|
| 624 |
+
"eval_samples_per_second": 31.282,
|
| 625 |
+
"eval_steps_per_second": 2.3,
|
| 626 |
+
"step": 168
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 28.333333333333332,
|
| 630 |
+
"grad_norm": 0.01064391154795885,
|
| 631 |
+
"learning_rate": 2.425925925925926e-05,
|
| 632 |
+
"loss": 0.0003,
|
| 633 |
+
"step": 170
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 29.0,
|
| 637 |
+
"eval_accuracy": 0.8970588235294118,
|
| 638 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 639 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 640 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 641 |
+
"eval_loss": 0.6319212913513184,
|
| 642 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 643 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 644 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 645 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 646 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 647 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 648 |
+
"eval_runtime": 2.1953,
|
| 649 |
+
"eval_samples_per_second": 30.976,
|
| 650 |
+
"eval_steps_per_second": 2.278,
|
| 651 |
+
"step": 174
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"epoch": 30.0,
|
| 655 |
+
"grad_norm": 0.006854104809463024,
|
| 656 |
+
"learning_rate": 2.240740740740741e-05,
|
| 657 |
+
"loss": 0.0002,
|
| 658 |
+
"step": 180
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"epoch": 30.0,
|
| 662 |
+
"eval_accuracy": 0.8970588235294118,
|
| 663 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 664 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 665 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 666 |
+
"eval_loss": 0.6456788182258606,
|
| 667 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 668 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 669 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 670 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 671 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 672 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 673 |
+
"eval_runtime": 2.1711,
|
| 674 |
+
"eval_samples_per_second": 31.321,
|
| 675 |
+
"eval_steps_per_second": 2.303,
|
| 676 |
+
"step": 180
|
| 677 |
+
},
|
| 678 |
+
{
|
| 679 |
+
"epoch": 31.0,
|
| 680 |
+
"eval_accuracy": 0.8970588235294118,
|
| 681 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 682 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 683 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 684 |
+
"eval_loss": 0.6529943346977234,
|
| 685 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 686 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 687 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 688 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 689 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 690 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 691 |
+
"eval_runtime": 2.2044,
|
| 692 |
+
"eval_samples_per_second": 30.848,
|
| 693 |
+
"eval_steps_per_second": 2.268,
|
| 694 |
+
"step": 186
|
| 695 |
+
},
|
| 696 |
+
{
|
| 697 |
+
"epoch": 31.666666666666668,
|
| 698 |
+
"grad_norm": 0.006000952795147896,
|
| 699 |
+
"learning_rate": 2.0555555555555555e-05,
|
| 700 |
+
"loss": 0.0002,
|
| 701 |
+
"step": 190
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 32.0,
|
| 705 |
+
"eval_accuracy": 0.8970588235294118,
|
| 706 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 707 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 708 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 709 |
+
"eval_loss": 0.6579628586769104,
|
| 710 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 711 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 712 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 713 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 714 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 715 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 716 |
+
"eval_runtime": 2.1932,
|
| 717 |
+
"eval_samples_per_second": 31.005,
|
| 718 |
+
"eval_steps_per_second": 2.28,
|
| 719 |
+
"step": 192
|
| 720 |
+
},
|
| 721 |
+
{
|
| 722 |
+
"epoch": 33.0,
|
| 723 |
+
"eval_accuracy": 0.8970588235294118,
|
| 724 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 725 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 726 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 727 |
+
"eval_loss": 0.6649767160415649,
|
| 728 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 729 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 730 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 731 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 732 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 733 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 734 |
+
"eval_runtime": 2.1922,
|
| 735 |
+
"eval_samples_per_second": 31.019,
|
| 736 |
+
"eval_steps_per_second": 2.281,
|
| 737 |
+
"step": 198
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 33.333333333333336,
|
| 741 |
+
"grad_norm": 0.0073044863529503345,
|
| 742 |
+
"learning_rate": 1.8703703703703704e-05,
|
| 743 |
+
"loss": 0.0002,
|
| 744 |
+
"step": 200
|
| 745 |
+
},
|
| 746 |
+
{
|
| 747 |
+
"epoch": 34.0,
|
| 748 |
+
"eval_accuracy": 0.8970588235294118,
|
| 749 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 750 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 751 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 752 |
+
"eval_loss": 0.677439272403717,
|
| 753 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 754 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 755 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 756 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 757 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 758 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 759 |
+
"eval_runtime": 2.2369,
|
| 760 |
+
"eval_samples_per_second": 30.4,
|
| 761 |
+
"eval_steps_per_second": 2.235,
|
| 762 |
+
"step": 204
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 35.0,
|
| 766 |
+
"grad_norm": 0.00664471136406064,
|
| 767 |
+
"learning_rate": 1.6851851851851853e-05,
|
| 768 |
+
"loss": 0.0002,
|
| 769 |
+
"step": 210
|
| 770 |
+
},
|
| 771 |
+
{
|
| 772 |
+
"epoch": 35.0,
|
| 773 |
+
"eval_accuracy": 0.8970588235294118,
|
| 774 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 775 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 776 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 777 |
+
"eval_loss": 0.6827277541160583,
|
| 778 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 779 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 780 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 781 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 782 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 783 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 784 |
+
"eval_runtime": 2.2152,
|
| 785 |
+
"eval_samples_per_second": 30.697,
|
| 786 |
+
"eval_steps_per_second": 2.257,
|
| 787 |
+
"step": 210
|
| 788 |
+
},
|
| 789 |
+
{
|
| 790 |
+
"epoch": 36.0,
|
| 791 |
+
"eval_accuracy": 0.8970588235294118,
|
| 792 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 793 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 794 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 795 |
+
"eval_loss": 0.6850156188011169,
|
| 796 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 797 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 798 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 799 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 800 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 801 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 802 |
+
"eval_runtime": 2.1962,
|
| 803 |
+
"eval_samples_per_second": 30.963,
|
| 804 |
+
"eval_steps_per_second": 2.277,
|
| 805 |
+
"step": 216
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"epoch": 36.666666666666664,
|
| 809 |
+
"grad_norm": 0.003711260389536619,
|
| 810 |
+
"learning_rate": 1.5e-05,
|
| 811 |
+
"loss": 0.0001,
|
| 812 |
+
"step": 220
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"epoch": 37.0,
|
| 816 |
+
"eval_accuracy": 0.8970588235294118,
|
| 817 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 818 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 819 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 820 |
+
"eval_loss": 0.684583842754364,
|
| 821 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 822 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 823 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 824 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 825 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 826 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 827 |
+
"eval_runtime": 2.2135,
|
| 828 |
+
"eval_samples_per_second": 30.721,
|
| 829 |
+
"eval_steps_per_second": 2.259,
|
| 830 |
+
"step": 222
|
| 831 |
+
},
|
| 832 |
+
{
|
| 833 |
+
"epoch": 38.0,
|
| 834 |
+
"eval_accuracy": 0.8970588235294118,
|
| 835 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 836 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 837 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 838 |
+
"eval_loss": 0.6853307485580444,
|
| 839 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 840 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 841 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 842 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 843 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 844 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 845 |
+
"eval_runtime": 2.2406,
|
| 846 |
+
"eval_samples_per_second": 30.348,
|
| 847 |
+
"eval_steps_per_second": 2.231,
|
| 848 |
+
"step": 228
|
| 849 |
+
},
|
| 850 |
+
{
|
| 851 |
+
"epoch": 38.333333333333336,
|
| 852 |
+
"grad_norm": 0.003823125036433339,
|
| 853 |
+
"learning_rate": 1.3148148148148148e-05,
|
| 854 |
+
"loss": 0.0001,
|
| 855 |
+
"step": 230
|
| 856 |
+
},
|
| 857 |
+
{
|
| 858 |
+
"epoch": 39.0,
|
| 859 |
+
"eval_accuracy": 0.8970588235294118,
|
| 860 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 861 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 862 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 863 |
+
"eval_loss": 0.6853978633880615,
|
| 864 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 865 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 866 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 867 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 868 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 869 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 870 |
+
"eval_runtime": 2.1744,
|
| 871 |
+
"eval_samples_per_second": 31.273,
|
| 872 |
+
"eval_steps_per_second": 2.299,
|
| 873 |
+
"step": 234
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 40.0,
|
| 877 |
+
"grad_norm": 0.0067051672376692295,
|
| 878 |
+
"learning_rate": 1.1296296296296297e-05,
|
| 879 |
+
"loss": 0.0001,
|
| 880 |
+
"step": 240
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 40.0,
|
| 884 |
+
"eval_accuracy": 0.8970588235294118,
|
| 885 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 886 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 887 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 888 |
+
"eval_loss": 0.6844778656959534,
|
| 889 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 890 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 891 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 892 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 893 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 894 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 895 |
+
"eval_runtime": 2.2084,
|
| 896 |
+
"eval_samples_per_second": 30.791,
|
| 897 |
+
"eval_steps_per_second": 2.264,
|
| 898 |
+
"step": 240
|
| 899 |
+
},
|
| 900 |
+
{
|
| 901 |
+
"epoch": 41.0,
|
| 902 |
+
"eval_accuracy": 0.8970588235294118,
|
| 903 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 904 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 905 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 906 |
+
"eval_loss": 0.6846556663513184,
|
| 907 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 908 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 909 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 910 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 911 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 912 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 913 |
+
"eval_runtime": 2.2966,
|
| 914 |
+
"eval_samples_per_second": 29.61,
|
| 915 |
+
"eval_steps_per_second": 2.177,
|
| 916 |
+
"step": 246
|
| 917 |
+
},
|
| 918 |
+
{
|
| 919 |
+
"epoch": 41.666666666666664,
|
| 920 |
+
"grad_norm": 0.005940043367445469,
|
| 921 |
+
"learning_rate": 9.444444444444445e-06,
|
| 922 |
+
"loss": 0.0001,
|
| 923 |
+
"step": 250
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 42.0,
|
| 927 |
+
"eval_accuracy": 0.8970588235294118,
|
| 928 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 929 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 930 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 931 |
+
"eval_loss": 0.6859800815582275,
|
| 932 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 933 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 934 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 935 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 936 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 937 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 938 |
+
"eval_runtime": 2.2361,
|
| 939 |
+
"eval_samples_per_second": 30.41,
|
| 940 |
+
"eval_steps_per_second": 2.236,
|
| 941 |
+
"step": 252
|
| 942 |
+
},
|
| 943 |
+
{
|
| 944 |
+
"epoch": 43.0,
|
| 945 |
+
"eval_accuracy": 0.8970588235294118,
|
| 946 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 947 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 948 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 949 |
+
"eval_loss": 0.6874513030052185,
|
| 950 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 951 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 952 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 953 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 954 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 955 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 956 |
+
"eval_runtime": 2.1797,
|
| 957 |
+
"eval_samples_per_second": 31.196,
|
| 958 |
+
"eval_steps_per_second": 2.294,
|
| 959 |
+
"step": 258
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 43.333333333333336,
|
| 963 |
+
"grad_norm": 0.003746297210454941,
|
| 964 |
+
"learning_rate": 7.592592592592593e-06,
|
| 965 |
+
"loss": 0.0001,
|
| 966 |
+
"step": 260
|
| 967 |
+
},
|
| 968 |
+
{
|
| 969 |
+
"epoch": 44.0,
|
| 970 |
+
"eval_accuracy": 0.8970588235294118,
|
| 971 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 972 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 973 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 974 |
+
"eval_loss": 0.6894810199737549,
|
| 975 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 976 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 977 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 978 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 979 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 980 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 981 |
+
"eval_runtime": 2.2822,
|
| 982 |
+
"eval_samples_per_second": 29.796,
|
| 983 |
+
"eval_steps_per_second": 2.191,
|
| 984 |
+
"step": 264
|
| 985 |
+
},
|
| 986 |
+
{
|
| 987 |
+
"epoch": 45.0,
|
| 988 |
+
"grad_norm": 0.010260913521051407,
|
| 989 |
+
"learning_rate": 5.740740740740741e-06,
|
| 990 |
+
"loss": 0.0001,
|
| 991 |
+
"step": 270
|
| 992 |
+
},
|
| 993 |
+
{
|
| 994 |
+
"epoch": 45.0,
|
| 995 |
+
"eval_accuracy": 0.8970588235294118,
|
| 996 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 997 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 998 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 999 |
+
"eval_loss": 0.6902977824211121,
|
| 1000 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 1001 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 1002 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 1003 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 1004 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 1005 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 1006 |
+
"eval_runtime": 2.2179,
|
| 1007 |
+
"eval_samples_per_second": 30.66,
|
| 1008 |
+
"eval_steps_per_second": 2.254,
|
| 1009 |
+
"step": 270
|
| 1010 |
+
},
|
| 1011 |
+
{
|
| 1012 |
+
"epoch": 46.0,
|
| 1013 |
+
"eval_accuracy": 0.8970588235294118,
|
| 1014 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 1015 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 1016 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 1017 |
+
"eval_loss": 0.691429078578949,
|
| 1018 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 1019 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 1020 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 1021 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 1022 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 1023 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 1024 |
+
"eval_runtime": 2.2021,
|
| 1025 |
+
"eval_samples_per_second": 30.88,
|
| 1026 |
+
"eval_steps_per_second": 2.271,
|
| 1027 |
+
"step": 276
|
| 1028 |
+
},
|
| 1029 |
+
{
|
| 1030 |
+
"epoch": 46.666666666666664,
|
| 1031 |
+
"grad_norm": 0.003823869628831744,
|
| 1032 |
+
"learning_rate": 3.888888888888889e-06,
|
| 1033 |
+
"loss": 0.0001,
|
| 1034 |
+
"step": 280
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 47.0,
|
| 1038 |
+
"eval_accuracy": 0.8970588235294118,
|
| 1039 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 1040 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 1041 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 1042 |
+
"eval_loss": 0.6919637322425842,
|
| 1043 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 1044 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 1045 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 1046 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 1047 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 1048 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 1049 |
+
"eval_runtime": 2.1849,
|
| 1050 |
+
"eval_samples_per_second": 31.123,
|
| 1051 |
+
"eval_steps_per_second": 2.288,
|
| 1052 |
+
"step": 282
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 48.0,
|
| 1056 |
+
"eval_accuracy": 0.8970588235294118,
|
| 1057 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 1058 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 1059 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 1060 |
+
"eval_loss": 0.6921000480651855,
|
| 1061 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 1062 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 1063 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 1064 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 1065 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 1066 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 1067 |
+
"eval_runtime": 2.1918,
|
| 1068 |
+
"eval_samples_per_second": 31.025,
|
| 1069 |
+
"eval_steps_per_second": 2.281,
|
| 1070 |
+
"step": 288
|
| 1071 |
+
},
|
| 1072 |
+
{
|
| 1073 |
+
"epoch": 48.333333333333336,
|
| 1074 |
+
"grad_norm": 0.004158890340477228,
|
| 1075 |
+
"learning_rate": 2.0370370370370375e-06,
|
| 1076 |
+
"loss": 0.0001,
|
| 1077 |
+
"step": 290
|
| 1078 |
+
},
|
| 1079 |
+
{
|
| 1080 |
+
"epoch": 49.0,
|
| 1081 |
+
"eval_accuracy": 0.8970588235294118,
|
| 1082 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 1083 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 1084 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 1085 |
+
"eval_loss": 0.6922561526298523,
|
| 1086 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 1087 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 1088 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 1089 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 1090 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 1091 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 1092 |
+
"eval_runtime": 2.2198,
|
| 1093 |
+
"eval_samples_per_second": 30.634,
|
| 1094 |
+
"eval_steps_per_second": 2.252,
|
| 1095 |
+
"step": 294
|
| 1096 |
+
},
|
| 1097 |
+
{
|
| 1098 |
+
"epoch": 50.0,
|
| 1099 |
+
"grad_norm": 0.003357668872922659,
|
| 1100 |
+
"learning_rate": 1.851851851851852e-07,
|
| 1101 |
+
"loss": 0.0001,
|
| 1102 |
+
"step": 300
|
| 1103 |
+
},
|
| 1104 |
+
{
|
| 1105 |
+
"epoch": 50.0,
|
| 1106 |
+
"eval_accuracy": 0.8970588235294118,
|
| 1107 |
+
"eval_f1_dry": 0.8888888888888888,
|
| 1108 |
+
"eval_f1_humid": 0.9041095890410958,
|
| 1109 |
+
"eval_f1_macro": 0.8964992389649924,
|
| 1110 |
+
"eval_loss": 0.692406177520752,
|
| 1111 |
+
"eval_precision_dry": 0.8235294117647058,
|
| 1112 |
+
"eval_precision_humid": 0.9705882352941176,
|
| 1113 |
+
"eval_precision_macro": 0.8970588235294117,
|
| 1114 |
+
"eval_recall_dry": 0.9655172413793104,
|
| 1115 |
+
"eval_recall_humid": 0.8461538461538461,
|
| 1116 |
+
"eval_recall_macro": 0.9058355437665783,
|
| 1117 |
+
"eval_runtime": 2.2412,
|
| 1118 |
+
"eval_samples_per_second": 30.341,
|
| 1119 |
+
"eval_steps_per_second": 2.231,
|
| 1120 |
+
"step": 300
|
| 1121 |
+
},
|
| 1122 |
+
{
|
| 1123 |
+
"epoch": 50.0,
|
| 1124 |
+
"step": 300,
|
| 1125 |
+
"total_flos": 4.3301918339334144e+18,
|
| 1126 |
+
"train_loss": 0.04831398472228708,
|
| 1127 |
+
"train_runtime": 1645.2014,
|
| 1128 |
+
"train_samples_per_second": 11.518,
|
| 1129 |
+
"train_steps_per_second": 0.182
|
| 1130 |
+
}
|
| 1131 |
+
],
|
| 1132 |
+
"logging_steps": 10,
|
| 1133 |
+
"max_steps": 300,
|
| 1134 |
+
"num_input_tokens_seen": 0,
|
| 1135 |
+
"num_train_epochs": 50,
|
| 1136 |
+
"save_steps": 500,
|
| 1137 |
+
"stateful_callbacks": {
|
| 1138 |
+
"TrainerControl": {
|
| 1139 |
+
"args": {
|
| 1140 |
+
"should_epoch_stop": false,
|
| 1141 |
+
"should_evaluate": false,
|
| 1142 |
+
"should_log": false,
|
| 1143 |
+
"should_save": true,
|
| 1144 |
+
"should_training_stop": true
|
| 1145 |
+
},
|
| 1146 |
+
"attributes": {}
|
| 1147 |
+
}
|
| 1148 |
+
},
|
| 1149 |
+
"total_flos": 4.3301918339334144e+18,
|
| 1150 |
+
"train_batch_size": 16,
|
| 1151 |
+
"trial_name": null,
|
| 1152 |
+
"trial_params": null
|
| 1153 |
+
}
|