Training in progress, epoch 0
Browse files- all_results.json +15 -0
- eval_results.json +10 -0
- pytorch_model.bin +1 -1
- train_results.json +7 -0
- trainer_state.json +4588 -0
- training_args.bin +1 -1
all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 9.99,
|
| 3 |
+
"eval_accuracy": 0.9322864572914583,
|
| 4 |
+
"eval_f1": 0.9324476370669043,
|
| 5 |
+
"eval_loss": 0.2586997151374817,
|
| 6 |
+
"eval_precision": 0.9332162039811837,
|
| 7 |
+
"eval_recall": 0.93260933633141,
|
| 8 |
+
"eval_runtime": 191.4658,
|
| 9 |
+
"eval_samples_per_second": 52.218,
|
| 10 |
+
"eval_steps_per_second": 1.938,
|
| 11 |
+
"train_loss": 0.7571127768142804,
|
| 12 |
+
"train_runtime": 37998.3887,
|
| 13 |
+
"train_samples_per_second": 21.049,
|
| 14 |
+
"train_steps_per_second": 0.195
|
| 15 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"eval_accuracy": 0.9322864572914583,
|
| 3 |
+
"eval_f1": 0.9324476370669043,
|
| 4 |
+
"eval_loss": 0.2586997151374817,
|
| 5 |
+
"eval_precision": 0.9332162039811837,
|
| 6 |
+
"eval_recall": 0.93260933633141,
|
| 7 |
+
"eval_runtime": 191.4658,
|
| 8 |
+
"eval_samples_per_second": 52.218,
|
| 9 |
+
"eval_steps_per_second": 1.938
|
| 10 |
+
}
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 351313597
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f6a0a3fe60e6393bdc54eaa3c4cf146b0a899365a4c10112e2eff62e06b40cc2
|
| 3 |
size 351313597
|
train_results.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 9.99,
|
| 3 |
+
"train_loss": 0.7571127768142804,
|
| 4 |
+
"train_runtime": 37998.3887,
|
| 5 |
+
"train_samples_per_second": 21.049,
|
| 6 |
+
"train_steps_per_second": 0.195
|
| 7 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,4588 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.9310862172434486,
|
| 3 |
+
"best_model_checkpoint": "10-convnextv2-base-22k-384-finetuned-spiderTraining100-1000/checkpoint-7400",
|
| 4 |
+
"epoch": 9.989875126560918,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 7400,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.01,
|
| 13 |
+
"learning_rate": 6.7567567567567575e-06,
|
| 14 |
+
"loss": 4.7191,
|
| 15 |
+
"step": 10
|
| 16 |
+
},
|
| 17 |
+
{
|
| 18 |
+
"epoch": 0.03,
|
| 19 |
+
"learning_rate": 1.3513513513513515e-05,
|
| 20 |
+
"loss": 4.6815,
|
| 21 |
+
"step": 20
|
| 22 |
+
},
|
| 23 |
+
{
|
| 24 |
+
"epoch": 0.04,
|
| 25 |
+
"learning_rate": 2.0270270270270273e-05,
|
| 26 |
+
"loss": 4.6441,
|
| 27 |
+
"step": 30
|
| 28 |
+
},
|
| 29 |
+
{
|
| 30 |
+
"epoch": 0.05,
|
| 31 |
+
"learning_rate": 2.702702702702703e-05,
|
| 32 |
+
"loss": 4.5566,
|
| 33 |
+
"step": 40
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"epoch": 0.07,
|
| 37 |
+
"learning_rate": 3.3783783783783784e-05,
|
| 38 |
+
"loss": 4.5796,
|
| 39 |
+
"step": 50
|
| 40 |
+
},
|
| 41 |
+
{
|
| 42 |
+
"epoch": 0.08,
|
| 43 |
+
"learning_rate": 4.0540540540540545e-05,
|
| 44 |
+
"loss": 4.6235,
|
| 45 |
+
"step": 60
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"epoch": 0.09,
|
| 49 |
+
"learning_rate": 4.72972972972973e-05,
|
| 50 |
+
"loss": 4.61,
|
| 51 |
+
"step": 70
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.11,
|
| 55 |
+
"learning_rate": 5.405405405405406e-05,
|
| 56 |
+
"loss": 4.6244,
|
| 57 |
+
"step": 80
|
| 58 |
+
},
|
| 59 |
+
{
|
| 60 |
+
"epoch": 0.12,
|
| 61 |
+
"learning_rate": 6.0810810810810814e-05,
|
| 62 |
+
"loss": 4.6277,
|
| 63 |
+
"step": 90
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"epoch": 0.13,
|
| 67 |
+
"learning_rate": 6.756756756756757e-05,
|
| 68 |
+
"loss": 4.6237,
|
| 69 |
+
"step": 100
|
| 70 |
+
},
|
| 71 |
+
{
|
| 72 |
+
"epoch": 0.15,
|
| 73 |
+
"learning_rate": 7.432432432432433e-05,
|
| 74 |
+
"loss": 4.5135,
|
| 75 |
+
"step": 110
|
| 76 |
+
},
|
| 77 |
+
{
|
| 78 |
+
"epoch": 0.16,
|
| 79 |
+
"learning_rate": 8.108108108108109e-05,
|
| 80 |
+
"loss": 4.4143,
|
| 81 |
+
"step": 120
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 0.18,
|
| 85 |
+
"learning_rate": 8.783783783783784e-05,
|
| 86 |
+
"loss": 4.1232,
|
| 87 |
+
"step": 130
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"epoch": 0.19,
|
| 91 |
+
"learning_rate": 9.45945945945946e-05,
|
| 92 |
+
"loss": 3.7123,
|
| 93 |
+
"step": 140
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.2,
|
| 97 |
+
"learning_rate": 0.00010135135135135136,
|
| 98 |
+
"loss": 3.1878,
|
| 99 |
+
"step": 150
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 0.22,
|
| 103 |
+
"learning_rate": 0.00010810810810810812,
|
| 104 |
+
"loss": 2.7711,
|
| 105 |
+
"step": 160
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"epoch": 0.23,
|
| 109 |
+
"learning_rate": 0.00011486486486486487,
|
| 110 |
+
"loss": 2.3975,
|
| 111 |
+
"step": 170
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"epoch": 0.24,
|
| 115 |
+
"learning_rate": 0.00012162162162162163,
|
| 116 |
+
"loss": 2.2354,
|
| 117 |
+
"step": 180
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"epoch": 0.26,
|
| 121 |
+
"learning_rate": 0.00012837837837837836,
|
| 122 |
+
"loss": 2.0567,
|
| 123 |
+
"step": 190
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 0.27,
|
| 127 |
+
"learning_rate": 0.00013513513513513514,
|
| 128 |
+
"loss": 1.9001,
|
| 129 |
+
"step": 200
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"epoch": 0.28,
|
| 133 |
+
"learning_rate": 0.00014189189189189188,
|
| 134 |
+
"loss": 1.7957,
|
| 135 |
+
"step": 210
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.3,
|
| 139 |
+
"learning_rate": 0.00014864864864864866,
|
| 140 |
+
"loss": 1.683,
|
| 141 |
+
"step": 220
|
| 142 |
+
},
|
| 143 |
+
{
|
| 144 |
+
"epoch": 0.31,
|
| 145 |
+
"learning_rate": 0.0001554054054054054,
|
| 146 |
+
"loss": 1.6268,
|
| 147 |
+
"step": 230
|
| 148 |
+
},
|
| 149 |
+
{
|
| 150 |
+
"epoch": 0.32,
|
| 151 |
+
"learning_rate": 0.00016216216216216218,
|
| 152 |
+
"loss": 1.6159,
|
| 153 |
+
"step": 240
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"epoch": 0.34,
|
| 157 |
+
"learning_rate": 0.00016891891891891893,
|
| 158 |
+
"loss": 1.4989,
|
| 159 |
+
"step": 250
|
| 160 |
+
},
|
| 161 |
+
{
|
| 162 |
+
"epoch": 0.35,
|
| 163 |
+
"learning_rate": 0.00017567567567567568,
|
| 164 |
+
"loss": 1.4819,
|
| 165 |
+
"step": 260
|
| 166 |
+
},
|
| 167 |
+
{
|
| 168 |
+
"epoch": 0.36,
|
| 169 |
+
"learning_rate": 0.00018243243243243242,
|
| 170 |
+
"loss": 1.4353,
|
| 171 |
+
"step": 270
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"epoch": 0.38,
|
| 175 |
+
"learning_rate": 0.0001891891891891892,
|
| 176 |
+
"loss": 1.4604,
|
| 177 |
+
"step": 280
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 0.39,
|
| 181 |
+
"learning_rate": 0.00019594594594594594,
|
| 182 |
+
"loss": 1.4115,
|
| 183 |
+
"step": 290
|
| 184 |
+
},
|
| 185 |
+
{
|
| 186 |
+
"epoch": 0.4,
|
| 187 |
+
"learning_rate": 0.00020270270270270272,
|
| 188 |
+
"loss": 1.3756,
|
| 189 |
+
"step": 300
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 0.42,
|
| 193 |
+
"learning_rate": 0.00020945945945945947,
|
| 194 |
+
"loss": 1.3704,
|
| 195 |
+
"step": 310
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 0.43,
|
| 199 |
+
"learning_rate": 0.00021621621621621624,
|
| 200 |
+
"loss": 1.3525,
|
| 201 |
+
"step": 320
|
| 202 |
+
},
|
| 203 |
+
{
|
| 204 |
+
"epoch": 0.45,
|
| 205 |
+
"learning_rate": 0.000222972972972973,
|
| 206 |
+
"loss": 1.3842,
|
| 207 |
+
"step": 330
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"epoch": 0.46,
|
| 211 |
+
"learning_rate": 0.00022972972972972974,
|
| 212 |
+
"loss": 1.3297,
|
| 213 |
+
"step": 340
|
| 214 |
+
},
|
| 215 |
+
{
|
| 216 |
+
"epoch": 0.47,
|
| 217 |
+
"learning_rate": 0.00023648648648648648,
|
| 218 |
+
"loss": 1.304,
|
| 219 |
+
"step": 350
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 0.49,
|
| 223 |
+
"learning_rate": 0.00024324324324324326,
|
| 224 |
+
"loss": 1.4385,
|
| 225 |
+
"step": 360
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"epoch": 0.5,
|
| 229 |
+
"learning_rate": 0.00025,
|
| 230 |
+
"loss": 1.3085,
|
| 231 |
+
"step": 370
|
| 232 |
+
},
|
| 233 |
+
{
|
| 234 |
+
"epoch": 0.51,
|
| 235 |
+
"learning_rate": 0.0002567567567567567,
|
| 236 |
+
"loss": 1.3235,
|
| 237 |
+
"step": 380
|
| 238 |
+
},
|
| 239 |
+
{
|
| 240 |
+
"epoch": 0.53,
|
| 241 |
+
"learning_rate": 0.0002635135135135135,
|
| 242 |
+
"loss": 1.3504,
|
| 243 |
+
"step": 390
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"epoch": 0.54,
|
| 247 |
+
"learning_rate": 0.0002702702702702703,
|
| 248 |
+
"loss": 1.3193,
|
| 249 |
+
"step": 400
|
| 250 |
+
},
|
| 251 |
+
{
|
| 252 |
+
"epoch": 0.55,
|
| 253 |
+
"learning_rate": 0.00027702702702702705,
|
| 254 |
+
"loss": 1.2816,
|
| 255 |
+
"step": 410
|
| 256 |
+
},
|
| 257 |
+
{
|
| 258 |
+
"epoch": 0.57,
|
| 259 |
+
"learning_rate": 0.00028378378378378377,
|
| 260 |
+
"loss": 1.3683,
|
| 261 |
+
"step": 420
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 0.58,
|
| 265 |
+
"learning_rate": 0.00029054054054054054,
|
| 266 |
+
"loss": 1.1493,
|
| 267 |
+
"step": 430
|
| 268 |
+
},
|
| 269 |
+
{
|
| 270 |
+
"epoch": 0.59,
|
| 271 |
+
"learning_rate": 0.0002972972972972973,
|
| 272 |
+
"loss": 1.3138,
|
| 273 |
+
"step": 440
|
| 274 |
+
},
|
| 275 |
+
{
|
| 276 |
+
"epoch": 0.61,
|
| 277 |
+
"learning_rate": 0.00030405405405405404,
|
| 278 |
+
"loss": 1.3149,
|
| 279 |
+
"step": 450
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"epoch": 0.62,
|
| 283 |
+
"learning_rate": 0.0003108108108108108,
|
| 284 |
+
"loss": 1.297,
|
| 285 |
+
"step": 460
|
| 286 |
+
},
|
| 287 |
+
{
|
| 288 |
+
"epoch": 0.63,
|
| 289 |
+
"learning_rate": 0.00031756756756756753,
|
| 290 |
+
"loss": 1.4255,
|
| 291 |
+
"step": 470
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 0.65,
|
| 295 |
+
"learning_rate": 0.00032432432432432436,
|
| 296 |
+
"loss": 1.3951,
|
| 297 |
+
"step": 480
|
| 298 |
+
},
|
| 299 |
+
{
|
| 300 |
+
"epoch": 0.66,
|
| 301 |
+
"learning_rate": 0.0003310810810810811,
|
| 302 |
+
"loss": 1.2615,
|
| 303 |
+
"step": 490
|
| 304 |
+
},
|
| 305 |
+
{
|
| 306 |
+
"epoch": 0.67,
|
| 307 |
+
"learning_rate": 0.00033783783783783786,
|
| 308 |
+
"loss": 1.2754,
|
| 309 |
+
"step": 500
|
| 310 |
+
},
|
| 311 |
+
{
|
| 312 |
+
"epoch": 0.69,
|
| 313 |
+
"learning_rate": 0.0003445945945945946,
|
| 314 |
+
"loss": 1.3006,
|
| 315 |
+
"step": 510
|
| 316 |
+
},
|
| 317 |
+
{
|
| 318 |
+
"epoch": 0.7,
|
| 319 |
+
"learning_rate": 0.00035135135135135135,
|
| 320 |
+
"loss": 1.444,
|
| 321 |
+
"step": 520
|
| 322 |
+
},
|
| 323 |
+
{
|
| 324 |
+
"epoch": 0.72,
|
| 325 |
+
"learning_rate": 0.0003581081081081081,
|
| 326 |
+
"loss": 1.2068,
|
| 327 |
+
"step": 530
|
| 328 |
+
},
|
| 329 |
+
{
|
| 330 |
+
"epoch": 0.73,
|
| 331 |
+
"learning_rate": 0.00036486486486486485,
|
| 332 |
+
"loss": 1.4908,
|
| 333 |
+
"step": 540
|
| 334 |
+
},
|
| 335 |
+
{
|
| 336 |
+
"epoch": 0.74,
|
| 337 |
+
"learning_rate": 0.0003716216216216216,
|
| 338 |
+
"loss": 1.3093,
|
| 339 |
+
"step": 550
|
| 340 |
+
},
|
| 341 |
+
{
|
| 342 |
+
"epoch": 0.76,
|
| 343 |
+
"learning_rate": 0.0003783783783783784,
|
| 344 |
+
"loss": 1.2905,
|
| 345 |
+
"step": 560
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"epoch": 0.77,
|
| 349 |
+
"learning_rate": 0.00038513513513513517,
|
| 350 |
+
"loss": 1.4815,
|
| 351 |
+
"step": 570
|
| 352 |
+
},
|
| 353 |
+
{
|
| 354 |
+
"epoch": 0.78,
|
| 355 |
+
"learning_rate": 0.0003918918918918919,
|
| 356 |
+
"loss": 1.2824,
|
| 357 |
+
"step": 580
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"epoch": 0.8,
|
| 361 |
+
"learning_rate": 0.00039864864864864866,
|
| 362 |
+
"loss": 1.3966,
|
| 363 |
+
"step": 590
|
| 364 |
+
},
|
| 365 |
+
{
|
| 366 |
+
"epoch": 0.81,
|
| 367 |
+
"learning_rate": 0.00040540540540540544,
|
| 368 |
+
"loss": 1.3284,
|
| 369 |
+
"step": 600
|
| 370 |
+
},
|
| 371 |
+
{
|
| 372 |
+
"epoch": 0.82,
|
| 373 |
+
"learning_rate": 0.00041216216216216216,
|
| 374 |
+
"loss": 1.249,
|
| 375 |
+
"step": 610
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"epoch": 0.84,
|
| 379 |
+
"learning_rate": 0.00041891891891891893,
|
| 380 |
+
"loss": 1.2585,
|
| 381 |
+
"step": 620
|
| 382 |
+
},
|
| 383 |
+
{
|
| 384 |
+
"epoch": 0.85,
|
| 385 |
+
"learning_rate": 0.00042567567567567565,
|
| 386 |
+
"loss": 1.3343,
|
| 387 |
+
"step": 630
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 0.86,
|
| 391 |
+
"learning_rate": 0.0004324324324324325,
|
| 392 |
+
"loss": 1.2704,
|
| 393 |
+
"step": 640
|
| 394 |
+
},
|
| 395 |
+
{
|
| 396 |
+
"epoch": 0.88,
|
| 397 |
+
"learning_rate": 0.0004391891891891892,
|
| 398 |
+
"loss": 1.3528,
|
| 399 |
+
"step": 650
|
| 400 |
+
},
|
| 401 |
+
{
|
| 402 |
+
"epoch": 0.89,
|
| 403 |
+
"learning_rate": 0.000445945945945946,
|
| 404 |
+
"loss": 1.3853,
|
| 405 |
+
"step": 660
|
| 406 |
+
},
|
| 407 |
+
{
|
| 408 |
+
"epoch": 0.9,
|
| 409 |
+
"learning_rate": 0.0004527027027027027,
|
| 410 |
+
"loss": 1.3388,
|
| 411 |
+
"step": 670
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"epoch": 0.92,
|
| 415 |
+
"learning_rate": 0.00045945945945945947,
|
| 416 |
+
"loss": 1.4151,
|
| 417 |
+
"step": 680
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 0.93,
|
| 421 |
+
"learning_rate": 0.00046621621621621625,
|
| 422 |
+
"loss": 1.4184,
|
| 423 |
+
"step": 690
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"epoch": 0.94,
|
| 427 |
+
"learning_rate": 0.00047297297297297297,
|
| 428 |
+
"loss": 1.3309,
|
| 429 |
+
"step": 700
|
| 430 |
+
},
|
| 431 |
+
{
|
| 432 |
+
"epoch": 0.96,
|
| 433 |
+
"learning_rate": 0.00047972972972972974,
|
| 434 |
+
"loss": 1.4356,
|
| 435 |
+
"step": 710
|
| 436 |
+
},
|
| 437 |
+
{
|
| 438 |
+
"epoch": 0.97,
|
| 439 |
+
"learning_rate": 0.0004864864864864865,
|
| 440 |
+
"loss": 1.4564,
|
| 441 |
+
"step": 720
|
| 442 |
+
},
|
| 443 |
+
{
|
| 444 |
+
"epoch": 0.99,
|
| 445 |
+
"learning_rate": 0.0004932432432432432,
|
| 446 |
+
"loss": 1.3526,
|
| 447 |
+
"step": 730
|
| 448 |
+
},
|
| 449 |
+
{
|
| 450 |
+
"epoch": 1.0,
|
| 451 |
+
"learning_rate": 0.0005,
|
| 452 |
+
"loss": 1.2855,
|
| 453 |
+
"step": 740
|
| 454 |
+
},
|
| 455 |
+
{
|
| 456 |
+
"epoch": 1.0,
|
| 457 |
+
"eval_accuracy": 0.7249449889977996,
|
| 458 |
+
"eval_f1": 0.7228725416685348,
|
| 459 |
+
"eval_loss": 0.9570144414901733,
|
| 460 |
+
"eval_precision": 0.7636111347187294,
|
| 461 |
+
"eval_recall": 0.7236760047474455,
|
| 462 |
+
"eval_runtime": 196.5787,
|
| 463 |
+
"eval_samples_per_second": 50.86,
|
| 464 |
+
"eval_steps_per_second": 1.887,
|
| 465 |
+
"step": 740
|
| 466 |
+
},
|
| 467 |
+
{
|
| 468 |
+
"epoch": 1.01,
|
| 469 |
+
"learning_rate": 0.0004992492492492493,
|
| 470 |
+
"loss": 1.2199,
|
| 471 |
+
"step": 750
|
| 472 |
+
},
|
| 473 |
+
{
|
| 474 |
+
"epoch": 1.03,
|
| 475 |
+
"learning_rate": 0.0004984984984984984,
|
| 476 |
+
"loss": 1.3408,
|
| 477 |
+
"step": 760
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 1.04,
|
| 481 |
+
"learning_rate": 0.0004977477477477478,
|
| 482 |
+
"loss": 1.2037,
|
| 483 |
+
"step": 770
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"epoch": 1.05,
|
| 487 |
+
"learning_rate": 0.000496996996996997,
|
| 488 |
+
"loss": 1.3243,
|
| 489 |
+
"step": 780
|
| 490 |
+
},
|
| 491 |
+
{
|
| 492 |
+
"epoch": 1.07,
|
| 493 |
+
"learning_rate": 0.0004962462462462463,
|
| 494 |
+
"loss": 1.3089,
|
| 495 |
+
"step": 790
|
| 496 |
+
},
|
| 497 |
+
{
|
| 498 |
+
"epoch": 1.08,
|
| 499 |
+
"learning_rate": 0.0004954954954954955,
|
| 500 |
+
"loss": 1.2901,
|
| 501 |
+
"step": 800
|
| 502 |
+
},
|
| 503 |
+
{
|
| 504 |
+
"epoch": 1.09,
|
| 505 |
+
"learning_rate": 0.0004947447447447447,
|
| 506 |
+
"loss": 1.2378,
|
| 507 |
+
"step": 810
|
| 508 |
+
},
|
| 509 |
+
{
|
| 510 |
+
"epoch": 1.11,
|
| 511 |
+
"learning_rate": 0.0004939939939939941,
|
| 512 |
+
"loss": 1.2818,
|
| 513 |
+
"step": 820
|
| 514 |
+
},
|
| 515 |
+
{
|
| 516 |
+
"epoch": 1.12,
|
| 517 |
+
"learning_rate": 0.0004932432432432432,
|
| 518 |
+
"loss": 1.2249,
|
| 519 |
+
"step": 830
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"epoch": 1.13,
|
| 523 |
+
"learning_rate": 0.0004924924924924925,
|
| 524 |
+
"loss": 1.2199,
|
| 525 |
+
"step": 840
|
| 526 |
+
},
|
| 527 |
+
{
|
| 528 |
+
"epoch": 1.15,
|
| 529 |
+
"learning_rate": 0.0004917417417417418,
|
| 530 |
+
"loss": 1.2566,
|
| 531 |
+
"step": 850
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"epoch": 1.16,
|
| 535 |
+
"learning_rate": 0.0004909909909909909,
|
| 536 |
+
"loss": 1.2868,
|
| 537 |
+
"step": 860
|
| 538 |
+
},
|
| 539 |
+
{
|
| 540 |
+
"epoch": 1.17,
|
| 541 |
+
"learning_rate": 0.0004902402402402403,
|
| 542 |
+
"loss": 1.4444,
|
| 543 |
+
"step": 870
|
| 544 |
+
},
|
| 545 |
+
{
|
| 546 |
+
"epoch": 1.19,
|
| 547 |
+
"learning_rate": 0.0004894894894894895,
|
| 548 |
+
"loss": 1.3055,
|
| 549 |
+
"step": 880
|
| 550 |
+
},
|
| 551 |
+
{
|
| 552 |
+
"epoch": 1.2,
|
| 553 |
+
"learning_rate": 0.0004887387387387388,
|
| 554 |
+
"loss": 1.2765,
|
| 555 |
+
"step": 890
|
| 556 |
+
},
|
| 557 |
+
{
|
| 558 |
+
"epoch": 1.21,
|
| 559 |
+
"learning_rate": 0.000487987987987988,
|
| 560 |
+
"loss": 1.2846,
|
| 561 |
+
"step": 900
|
| 562 |
+
},
|
| 563 |
+
{
|
| 564 |
+
"epoch": 1.23,
|
| 565 |
+
"learning_rate": 0.00048723723723723724,
|
| 566 |
+
"loss": 1.2887,
|
| 567 |
+
"step": 910
|
| 568 |
+
},
|
| 569 |
+
{
|
| 570 |
+
"epoch": 1.24,
|
| 571 |
+
"learning_rate": 0.0004864864864864865,
|
| 572 |
+
"loss": 1.2308,
|
| 573 |
+
"step": 920
|
| 574 |
+
},
|
| 575 |
+
{
|
| 576 |
+
"epoch": 1.26,
|
| 577 |
+
"learning_rate": 0.0004857357357357358,
|
| 578 |
+
"loss": 1.2236,
|
| 579 |
+
"step": 930
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"epoch": 1.27,
|
| 583 |
+
"learning_rate": 0.000484984984984985,
|
| 584 |
+
"loss": 1.2052,
|
| 585 |
+
"step": 940
|
| 586 |
+
},
|
| 587 |
+
{
|
| 588 |
+
"epoch": 1.28,
|
| 589 |
+
"learning_rate": 0.0004842342342342342,
|
| 590 |
+
"loss": 1.1945,
|
| 591 |
+
"step": 950
|
| 592 |
+
},
|
| 593 |
+
{
|
| 594 |
+
"epoch": 1.3,
|
| 595 |
+
"learning_rate": 0.0004834834834834835,
|
| 596 |
+
"loss": 1.2301,
|
| 597 |
+
"step": 960
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"epoch": 1.31,
|
| 601 |
+
"learning_rate": 0.0004827327327327327,
|
| 602 |
+
"loss": 1.2469,
|
| 603 |
+
"step": 970
|
| 604 |
+
},
|
| 605 |
+
{
|
| 606 |
+
"epoch": 1.32,
|
| 607 |
+
"learning_rate": 0.00048198198198198204,
|
| 608 |
+
"loss": 1.1334,
|
| 609 |
+
"step": 980
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"epoch": 1.34,
|
| 613 |
+
"learning_rate": 0.00048123123123123125,
|
| 614 |
+
"loss": 1.2563,
|
| 615 |
+
"step": 990
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 1.35,
|
| 619 |
+
"learning_rate": 0.00048048048048048047,
|
| 620 |
+
"loss": 1.2526,
|
| 621 |
+
"step": 1000
|
| 622 |
+
},
|
| 623 |
+
{
|
| 624 |
+
"epoch": 1.36,
|
| 625 |
+
"learning_rate": 0.00047972972972972974,
|
| 626 |
+
"loss": 1.3089,
|
| 627 |
+
"step": 1010
|
| 628 |
+
},
|
| 629 |
+
{
|
| 630 |
+
"epoch": 1.38,
|
| 631 |
+
"learning_rate": 0.00047897897897897896,
|
| 632 |
+
"loss": 1.1471,
|
| 633 |
+
"step": 1020
|
| 634 |
+
},
|
| 635 |
+
{
|
| 636 |
+
"epoch": 1.39,
|
| 637 |
+
"learning_rate": 0.0004782282282282283,
|
| 638 |
+
"loss": 1.2159,
|
| 639 |
+
"step": 1030
|
| 640 |
+
},
|
| 641 |
+
{
|
| 642 |
+
"epoch": 1.4,
|
| 643 |
+
"learning_rate": 0.0004774774774774775,
|
| 644 |
+
"loss": 1.254,
|
| 645 |
+
"step": 1040
|
| 646 |
+
},
|
| 647 |
+
{
|
| 648 |
+
"epoch": 1.42,
|
| 649 |
+
"learning_rate": 0.0004767267267267267,
|
| 650 |
+
"loss": 1.1875,
|
| 651 |
+
"step": 1050
|
| 652 |
+
},
|
| 653 |
+
{
|
| 654 |
+
"epoch": 1.43,
|
| 655 |
+
"learning_rate": 0.000475975975975976,
|
| 656 |
+
"loss": 1.28,
|
| 657 |
+
"step": 1060
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 1.44,
|
| 661 |
+
"learning_rate": 0.0004752252252252252,
|
| 662 |
+
"loss": 1.2797,
|
| 663 |
+
"step": 1070
|
| 664 |
+
},
|
| 665 |
+
{
|
| 666 |
+
"epoch": 1.46,
|
| 667 |
+
"learning_rate": 0.0004744744744744745,
|
| 668 |
+
"loss": 1.2614,
|
| 669 |
+
"step": 1080
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"epoch": 1.47,
|
| 673 |
+
"learning_rate": 0.00047372372372372375,
|
| 674 |
+
"loss": 1.2121,
|
| 675 |
+
"step": 1090
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 1.48,
|
| 679 |
+
"learning_rate": 0.00047297297297297297,
|
| 680 |
+
"loss": 1.2221,
|
| 681 |
+
"step": 1100
|
| 682 |
+
},
|
| 683 |
+
{
|
| 684 |
+
"epoch": 1.5,
|
| 685 |
+
"learning_rate": 0.00047222222222222224,
|
| 686 |
+
"loss": 1.1854,
|
| 687 |
+
"step": 1110
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"epoch": 1.51,
|
| 691 |
+
"learning_rate": 0.00047147147147147146,
|
| 692 |
+
"loss": 1.147,
|
| 693 |
+
"step": 1120
|
| 694 |
+
},
|
| 695 |
+
{
|
| 696 |
+
"epoch": 1.53,
|
| 697 |
+
"learning_rate": 0.0004707207207207207,
|
| 698 |
+
"loss": 1.1233,
|
| 699 |
+
"step": 1130
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
"epoch": 1.54,
|
| 703 |
+
"learning_rate": 0.00046996996996997,
|
| 704 |
+
"loss": 1.1032,
|
| 705 |
+
"step": 1140
|
| 706 |
+
},
|
| 707 |
+
{
|
| 708 |
+
"epoch": 1.55,
|
| 709 |
+
"learning_rate": 0.0004692192192192192,
|
| 710 |
+
"loss": 1.1887,
|
| 711 |
+
"step": 1150
|
| 712 |
+
},
|
| 713 |
+
{
|
| 714 |
+
"epoch": 1.57,
|
| 715 |
+
"learning_rate": 0.0004684684684684685,
|
| 716 |
+
"loss": 1.1418,
|
| 717 |
+
"step": 1160
|
| 718 |
+
},
|
| 719 |
+
{
|
| 720 |
+
"epoch": 1.58,
|
| 721 |
+
"learning_rate": 0.0004677177177177177,
|
| 722 |
+
"loss": 1.283,
|
| 723 |
+
"step": 1170
|
| 724 |
+
},
|
| 725 |
+
{
|
| 726 |
+
"epoch": 1.59,
|
| 727 |
+
"learning_rate": 0.000466966966966967,
|
| 728 |
+
"loss": 1.1715,
|
| 729 |
+
"step": 1180
|
| 730 |
+
},
|
| 731 |
+
{
|
| 732 |
+
"epoch": 1.61,
|
| 733 |
+
"learning_rate": 0.00046621621621621625,
|
| 734 |
+
"loss": 1.1116,
|
| 735 |
+
"step": 1190
|
| 736 |
+
},
|
| 737 |
+
{
|
| 738 |
+
"epoch": 1.62,
|
| 739 |
+
"learning_rate": 0.00046546546546546546,
|
| 740 |
+
"loss": 1.1327,
|
| 741 |
+
"step": 1200
|
| 742 |
+
},
|
| 743 |
+
{
|
| 744 |
+
"epoch": 1.63,
|
| 745 |
+
"learning_rate": 0.00046471471471471473,
|
| 746 |
+
"loss": 1.1484,
|
| 747 |
+
"step": 1210
|
| 748 |
+
},
|
| 749 |
+
{
|
| 750 |
+
"epoch": 1.65,
|
| 751 |
+
"learning_rate": 0.00046396396396396395,
|
| 752 |
+
"loss": 1.0976,
|
| 753 |
+
"step": 1220
|
| 754 |
+
},
|
| 755 |
+
{
|
| 756 |
+
"epoch": 1.66,
|
| 757 |
+
"learning_rate": 0.0004632132132132132,
|
| 758 |
+
"loss": 1.2164,
|
| 759 |
+
"step": 1230
|
| 760 |
+
},
|
| 761 |
+
{
|
| 762 |
+
"epoch": 1.67,
|
| 763 |
+
"learning_rate": 0.0004624624624624625,
|
| 764 |
+
"loss": 1.108,
|
| 765 |
+
"step": 1240
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"epoch": 1.69,
|
| 769 |
+
"learning_rate": 0.0004617117117117117,
|
| 770 |
+
"loss": 1.0269,
|
| 771 |
+
"step": 1250
|
| 772 |
+
},
|
| 773 |
+
{
|
| 774 |
+
"epoch": 1.7,
|
| 775 |
+
"learning_rate": 0.000460960960960961,
|
| 776 |
+
"loss": 1.1285,
|
| 777 |
+
"step": 1260
|
| 778 |
+
},
|
| 779 |
+
{
|
| 780 |
+
"epoch": 1.71,
|
| 781 |
+
"learning_rate": 0.0004602102102102102,
|
| 782 |
+
"loss": 1.0542,
|
| 783 |
+
"step": 1270
|
| 784 |
+
},
|
| 785 |
+
{
|
| 786 |
+
"epoch": 1.73,
|
| 787 |
+
"learning_rate": 0.00045945945945945947,
|
| 788 |
+
"loss": 1.1624,
|
| 789 |
+
"step": 1280
|
| 790 |
+
},
|
| 791 |
+
{
|
| 792 |
+
"epoch": 1.74,
|
| 793 |
+
"learning_rate": 0.00045870870870870874,
|
| 794 |
+
"loss": 1.1616,
|
| 795 |
+
"step": 1290
|
| 796 |
+
},
|
| 797 |
+
{
|
| 798 |
+
"epoch": 1.75,
|
| 799 |
+
"learning_rate": 0.00045795795795795796,
|
| 800 |
+
"loss": 1.1036,
|
| 801 |
+
"step": 1300
|
| 802 |
+
},
|
| 803 |
+
{
|
| 804 |
+
"epoch": 1.77,
|
| 805 |
+
"learning_rate": 0.00045720720720720723,
|
| 806 |
+
"loss": 1.0506,
|
| 807 |
+
"step": 1310
|
| 808 |
+
},
|
| 809 |
+
{
|
| 810 |
+
"epoch": 1.78,
|
| 811 |
+
"learning_rate": 0.00045645645645645645,
|
| 812 |
+
"loss": 1.1882,
|
| 813 |
+
"step": 1320
|
| 814 |
+
},
|
| 815 |
+
{
|
| 816 |
+
"epoch": 1.8,
|
| 817 |
+
"learning_rate": 0.0004557057057057057,
|
| 818 |
+
"loss": 1.1138,
|
| 819 |
+
"step": 1330
|
| 820 |
+
},
|
| 821 |
+
{
|
| 822 |
+
"epoch": 1.81,
|
| 823 |
+
"learning_rate": 0.000454954954954955,
|
| 824 |
+
"loss": 1.1566,
|
| 825 |
+
"step": 1340
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"epoch": 1.82,
|
| 829 |
+
"learning_rate": 0.0004542042042042042,
|
| 830 |
+
"loss": 1.1016,
|
| 831 |
+
"step": 1350
|
| 832 |
+
},
|
| 833 |
+
{
|
| 834 |
+
"epoch": 1.84,
|
| 835 |
+
"learning_rate": 0.0004534534534534535,
|
| 836 |
+
"loss": 1.0219,
|
| 837 |
+
"step": 1360
|
| 838 |
+
},
|
| 839 |
+
{
|
| 840 |
+
"epoch": 1.85,
|
| 841 |
+
"learning_rate": 0.0004527027027027027,
|
| 842 |
+
"loss": 1.0964,
|
| 843 |
+
"step": 1370
|
| 844 |
+
},
|
| 845 |
+
{
|
| 846 |
+
"epoch": 1.86,
|
| 847 |
+
"learning_rate": 0.0004519519519519519,
|
| 848 |
+
"loss": 1.0864,
|
| 849 |
+
"step": 1380
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"epoch": 1.88,
|
| 853 |
+
"learning_rate": 0.00045120120120120124,
|
| 854 |
+
"loss": 1.15,
|
| 855 |
+
"step": 1390
|
| 856 |
+
},
|
| 857 |
+
{
|
| 858 |
+
"epoch": 1.89,
|
| 859 |
+
"learning_rate": 0.00045045045045045046,
|
| 860 |
+
"loss": 1.074,
|
| 861 |
+
"step": 1400
|
| 862 |
+
},
|
| 863 |
+
{
|
| 864 |
+
"epoch": 1.9,
|
| 865 |
+
"learning_rate": 0.00044969969969969973,
|
| 866 |
+
"loss": 0.9878,
|
| 867 |
+
"step": 1410
|
| 868 |
+
},
|
| 869 |
+
{
|
| 870 |
+
"epoch": 1.92,
|
| 871 |
+
"learning_rate": 0.00044894894894894895,
|
| 872 |
+
"loss": 1.1312,
|
| 873 |
+
"step": 1420
|
| 874 |
+
},
|
| 875 |
+
{
|
| 876 |
+
"epoch": 1.93,
|
| 877 |
+
"learning_rate": 0.00044819819819819816,
|
| 878 |
+
"loss": 1.0452,
|
| 879 |
+
"step": 1430
|
| 880 |
+
},
|
| 881 |
+
{
|
| 882 |
+
"epoch": 1.94,
|
| 883 |
+
"learning_rate": 0.0004474474474474475,
|
| 884 |
+
"loss": 1.1523,
|
| 885 |
+
"step": 1440
|
| 886 |
+
},
|
| 887 |
+
{
|
| 888 |
+
"epoch": 1.96,
|
| 889 |
+
"learning_rate": 0.0004466966966966967,
|
| 890 |
+
"loss": 1.0347,
|
| 891 |
+
"step": 1450
|
| 892 |
+
},
|
| 893 |
+
{
|
| 894 |
+
"epoch": 1.97,
|
| 895 |
+
"learning_rate": 0.000445945945945946,
|
| 896 |
+
"loss": 1.1487,
|
| 897 |
+
"step": 1460
|
| 898 |
+
},
|
| 899 |
+
{
|
| 900 |
+
"epoch": 1.98,
|
| 901 |
+
"learning_rate": 0.0004451951951951952,
|
| 902 |
+
"loss": 1.1014,
|
| 903 |
+
"step": 1470
|
| 904 |
+
},
|
| 905 |
+
{
|
| 906 |
+
"epoch": 2.0,
|
| 907 |
+
"learning_rate": 0.0004444444444444444,
|
| 908 |
+
"loss": 1.1598,
|
| 909 |
+
"step": 1480
|
| 910 |
+
},
|
| 911 |
+
{
|
| 912 |
+
"epoch": 2.0,
|
| 913 |
+
"eval_accuracy": 0.7715543108621724,
|
| 914 |
+
"eval_f1": 0.7705672329120885,
|
| 915 |
+
"eval_loss": 0.8161230087280273,
|
| 916 |
+
"eval_precision": 0.7970684748506285,
|
| 917 |
+
"eval_recall": 0.7687323574914616,
|
| 918 |
+
"eval_runtime": 192.2102,
|
| 919 |
+
"eval_samples_per_second": 52.016,
|
| 920 |
+
"eval_steps_per_second": 1.93,
|
| 921 |
+
"step": 1481
|
| 922 |
+
},
|
| 923 |
+
{
|
| 924 |
+
"epoch": 2.01,
|
| 925 |
+
"learning_rate": 0.00044369369369369374,
|
| 926 |
+
"loss": 1.0385,
|
| 927 |
+
"step": 1490
|
| 928 |
+
},
|
| 929 |
+
{
|
| 930 |
+
"epoch": 2.02,
|
| 931 |
+
"learning_rate": 0.00044294294294294295,
|
| 932 |
+
"loss": 0.9801,
|
| 933 |
+
"step": 1500
|
| 934 |
+
},
|
| 935 |
+
{
|
| 936 |
+
"epoch": 2.04,
|
| 937 |
+
"learning_rate": 0.0004421921921921922,
|
| 938 |
+
"loss": 0.9433,
|
| 939 |
+
"step": 1510
|
| 940 |
+
},
|
| 941 |
+
{
|
| 942 |
+
"epoch": 2.05,
|
| 943 |
+
"learning_rate": 0.00044144144144144144,
|
| 944 |
+
"loss": 0.9969,
|
| 945 |
+
"step": 1520
|
| 946 |
+
},
|
| 947 |
+
{
|
| 948 |
+
"epoch": 2.07,
|
| 949 |
+
"learning_rate": 0.00044069069069069066,
|
| 950 |
+
"loss": 1.0062,
|
| 951 |
+
"step": 1530
|
| 952 |
+
},
|
| 953 |
+
{
|
| 954 |
+
"epoch": 2.08,
|
| 955 |
+
"learning_rate": 0.00043993993993994,
|
| 956 |
+
"loss": 0.9232,
|
| 957 |
+
"step": 1540
|
| 958 |
+
},
|
| 959 |
+
{
|
| 960 |
+
"epoch": 2.09,
|
| 961 |
+
"learning_rate": 0.0004391891891891892,
|
| 962 |
+
"loss": 1.0435,
|
| 963 |
+
"step": 1550
|
| 964 |
+
},
|
| 965 |
+
{
|
| 966 |
+
"epoch": 2.11,
|
| 967 |
+
"learning_rate": 0.0004384384384384385,
|
| 968 |
+
"loss": 0.9815,
|
| 969 |
+
"step": 1560
|
| 970 |
+
},
|
| 971 |
+
{
|
| 972 |
+
"epoch": 2.12,
|
| 973 |
+
"learning_rate": 0.0004376876876876877,
|
| 974 |
+
"loss": 0.98,
|
| 975 |
+
"step": 1570
|
| 976 |
+
},
|
| 977 |
+
{
|
| 978 |
+
"epoch": 2.13,
|
| 979 |
+
"learning_rate": 0.0004369369369369369,
|
| 980 |
+
"loss": 0.9993,
|
| 981 |
+
"step": 1580
|
| 982 |
+
},
|
| 983 |
+
{
|
| 984 |
+
"epoch": 2.15,
|
| 985 |
+
"learning_rate": 0.00043618618618618623,
|
| 986 |
+
"loss": 0.9791,
|
| 987 |
+
"step": 1590
|
| 988 |
+
},
|
| 989 |
+
{
|
| 990 |
+
"epoch": 2.16,
|
| 991 |
+
"learning_rate": 0.00043543543543543545,
|
| 992 |
+
"loss": 0.9976,
|
| 993 |
+
"step": 1600
|
| 994 |
+
},
|
| 995 |
+
{
|
| 996 |
+
"epoch": 2.17,
|
| 997 |
+
"learning_rate": 0.00043468468468468467,
|
| 998 |
+
"loss": 0.9826,
|
| 999 |
+
"step": 1610
|
| 1000 |
+
},
|
| 1001 |
+
{
|
| 1002 |
+
"epoch": 2.19,
|
| 1003 |
+
"learning_rate": 0.00043393393393393394,
|
| 1004 |
+
"loss": 0.9983,
|
| 1005 |
+
"step": 1620
|
| 1006 |
+
},
|
| 1007 |
+
{
|
| 1008 |
+
"epoch": 2.2,
|
| 1009 |
+
"learning_rate": 0.00043318318318318316,
|
| 1010 |
+
"loss": 0.9786,
|
| 1011 |
+
"step": 1630
|
| 1012 |
+
},
|
| 1013 |
+
{
|
| 1014 |
+
"epoch": 2.21,
|
| 1015 |
+
"learning_rate": 0.0004324324324324325,
|
| 1016 |
+
"loss": 0.9759,
|
| 1017 |
+
"step": 1640
|
| 1018 |
+
},
|
| 1019 |
+
{
|
| 1020 |
+
"epoch": 2.23,
|
| 1021 |
+
"learning_rate": 0.0004316816816816817,
|
| 1022 |
+
"loss": 0.9812,
|
| 1023 |
+
"step": 1650
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"epoch": 2.24,
|
| 1027 |
+
"learning_rate": 0.0004309309309309309,
|
| 1028 |
+
"loss": 0.9019,
|
| 1029 |
+
"step": 1660
|
| 1030 |
+
},
|
| 1031 |
+
{
|
| 1032 |
+
"epoch": 2.25,
|
| 1033 |
+
"learning_rate": 0.0004301801801801802,
|
| 1034 |
+
"loss": 0.9509,
|
| 1035 |
+
"step": 1670
|
| 1036 |
+
},
|
| 1037 |
+
{
|
| 1038 |
+
"epoch": 2.27,
|
| 1039 |
+
"learning_rate": 0.0004294294294294294,
|
| 1040 |
+
"loss": 0.8406,
|
| 1041 |
+
"step": 1680
|
| 1042 |
+
},
|
| 1043 |
+
{
|
| 1044 |
+
"epoch": 2.28,
|
| 1045 |
+
"learning_rate": 0.00042867867867867873,
|
| 1046 |
+
"loss": 0.8635,
|
| 1047 |
+
"step": 1690
|
| 1048 |
+
},
|
| 1049 |
+
{
|
| 1050 |
+
"epoch": 2.29,
|
| 1051 |
+
"learning_rate": 0.00042792792792792795,
|
| 1052 |
+
"loss": 0.9476,
|
| 1053 |
+
"step": 1700
|
| 1054 |
+
},
|
| 1055 |
+
{
|
| 1056 |
+
"epoch": 2.31,
|
| 1057 |
+
"learning_rate": 0.00042717717717717717,
|
| 1058 |
+
"loss": 0.9808,
|
| 1059 |
+
"step": 1710
|
| 1060 |
+
},
|
| 1061 |
+
{
|
| 1062 |
+
"epoch": 2.32,
|
| 1063 |
+
"learning_rate": 0.00042642642642642644,
|
| 1064 |
+
"loss": 0.8811,
|
| 1065 |
+
"step": 1720
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 2.34,
|
| 1069 |
+
"learning_rate": 0.00042567567567567565,
|
| 1070 |
+
"loss": 0.9824,
|
| 1071 |
+
"step": 1730
|
| 1072 |
+
},
|
| 1073 |
+
{
|
| 1074 |
+
"epoch": 2.35,
|
| 1075 |
+
"learning_rate": 0.000424924924924925,
|
| 1076 |
+
"loss": 0.8877,
|
| 1077 |
+
"step": 1740
|
| 1078 |
+
},
|
| 1079 |
+
{
|
| 1080 |
+
"epoch": 2.36,
|
| 1081 |
+
"learning_rate": 0.0004241741741741742,
|
| 1082 |
+
"loss": 0.9078,
|
| 1083 |
+
"step": 1750
|
| 1084 |
+
},
|
| 1085 |
+
{
|
| 1086 |
+
"epoch": 2.38,
|
| 1087 |
+
"learning_rate": 0.0004234234234234234,
|
| 1088 |
+
"loss": 0.9928,
|
| 1089 |
+
"step": 1760
|
| 1090 |
+
},
|
| 1091 |
+
{
|
| 1092 |
+
"epoch": 2.39,
|
| 1093 |
+
"learning_rate": 0.0004226726726726727,
|
| 1094 |
+
"loss": 0.9896,
|
| 1095 |
+
"step": 1770
|
| 1096 |
+
},
|
| 1097 |
+
{
|
| 1098 |
+
"epoch": 2.4,
|
| 1099 |
+
"learning_rate": 0.0004219219219219219,
|
| 1100 |
+
"loss": 0.9167,
|
| 1101 |
+
"step": 1780
|
| 1102 |
+
},
|
| 1103 |
+
{
|
| 1104 |
+
"epoch": 2.42,
|
| 1105 |
+
"learning_rate": 0.0004211711711711712,
|
| 1106 |
+
"loss": 0.9614,
|
| 1107 |
+
"step": 1790
|
| 1108 |
+
},
|
| 1109 |
+
{
|
| 1110 |
+
"epoch": 2.43,
|
| 1111 |
+
"learning_rate": 0.00042042042042042044,
|
| 1112 |
+
"loss": 0.9728,
|
| 1113 |
+
"step": 1800
|
| 1114 |
+
},
|
| 1115 |
+
{
|
| 1116 |
+
"epoch": 2.44,
|
| 1117 |
+
"learning_rate": 0.00041966966966966966,
|
| 1118 |
+
"loss": 0.8913,
|
| 1119 |
+
"step": 1810
|
| 1120 |
+
},
|
| 1121 |
+
{
|
| 1122 |
+
"epoch": 2.46,
|
| 1123 |
+
"learning_rate": 0.00041891891891891893,
|
| 1124 |
+
"loss": 0.9239,
|
| 1125 |
+
"step": 1820
|
| 1126 |
+
},
|
| 1127 |
+
{
|
| 1128 |
+
"epoch": 2.47,
|
| 1129 |
+
"learning_rate": 0.00041816816816816815,
|
| 1130 |
+
"loss": 0.8981,
|
| 1131 |
+
"step": 1830
|
| 1132 |
+
},
|
| 1133 |
+
{
|
| 1134 |
+
"epoch": 2.48,
|
| 1135 |
+
"learning_rate": 0.0004174174174174174,
|
| 1136 |
+
"loss": 0.9159,
|
| 1137 |
+
"step": 1840
|
| 1138 |
+
},
|
| 1139 |
+
{
|
| 1140 |
+
"epoch": 2.5,
|
| 1141 |
+
"learning_rate": 0.0004166666666666667,
|
| 1142 |
+
"loss": 0.9693,
|
| 1143 |
+
"step": 1850
|
| 1144 |
+
},
|
| 1145 |
+
{
|
| 1146 |
+
"epoch": 2.51,
|
| 1147 |
+
"learning_rate": 0.0004159159159159159,
|
| 1148 |
+
"loss": 0.9409,
|
| 1149 |
+
"step": 1860
|
| 1150 |
+
},
|
| 1151 |
+
{
|
| 1152 |
+
"epoch": 2.52,
|
| 1153 |
+
"learning_rate": 0.0004151651651651652,
|
| 1154 |
+
"loss": 0.9313,
|
| 1155 |
+
"step": 1870
|
| 1156 |
+
},
|
| 1157 |
+
{
|
| 1158 |
+
"epoch": 2.54,
|
| 1159 |
+
"learning_rate": 0.0004144144144144144,
|
| 1160 |
+
"loss": 0.8637,
|
| 1161 |
+
"step": 1880
|
| 1162 |
+
},
|
| 1163 |
+
{
|
| 1164 |
+
"epoch": 2.55,
|
| 1165 |
+
"learning_rate": 0.00041366366366366367,
|
| 1166 |
+
"loss": 0.9116,
|
| 1167 |
+
"step": 1890
|
| 1168 |
+
},
|
| 1169 |
+
{
|
| 1170 |
+
"epoch": 2.56,
|
| 1171 |
+
"learning_rate": 0.00041291291291291294,
|
| 1172 |
+
"loss": 0.9368,
|
| 1173 |
+
"step": 1900
|
| 1174 |
+
},
|
| 1175 |
+
{
|
| 1176 |
+
"epoch": 2.58,
|
| 1177 |
+
"learning_rate": 0.00041216216216216216,
|
| 1178 |
+
"loss": 0.9111,
|
| 1179 |
+
"step": 1910
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"epoch": 2.59,
|
| 1183 |
+
"learning_rate": 0.00041141141141141143,
|
| 1184 |
+
"loss": 0.9446,
|
| 1185 |
+
"step": 1920
|
| 1186 |
+
},
|
| 1187 |
+
{
|
| 1188 |
+
"epoch": 2.61,
|
| 1189 |
+
"learning_rate": 0.00041066066066066065,
|
| 1190 |
+
"loss": 0.9185,
|
| 1191 |
+
"step": 1930
|
| 1192 |
+
},
|
| 1193 |
+
{
|
| 1194 |
+
"epoch": 2.62,
|
| 1195 |
+
"learning_rate": 0.0004099099099099099,
|
| 1196 |
+
"loss": 0.8664,
|
| 1197 |
+
"step": 1940
|
| 1198 |
+
},
|
| 1199 |
+
{
|
| 1200 |
+
"epoch": 2.63,
|
| 1201 |
+
"learning_rate": 0.0004091591591591592,
|
| 1202 |
+
"loss": 0.7811,
|
| 1203 |
+
"step": 1950
|
| 1204 |
+
},
|
| 1205 |
+
{
|
| 1206 |
+
"epoch": 2.65,
|
| 1207 |
+
"learning_rate": 0.0004084084084084084,
|
| 1208 |
+
"loss": 0.9265,
|
| 1209 |
+
"step": 1960
|
| 1210 |
+
},
|
| 1211 |
+
{
|
| 1212 |
+
"epoch": 2.66,
|
| 1213 |
+
"learning_rate": 0.0004076576576576577,
|
| 1214 |
+
"loss": 0.83,
|
| 1215 |
+
"step": 1970
|
| 1216 |
+
},
|
| 1217 |
+
{
|
| 1218 |
+
"epoch": 2.67,
|
| 1219 |
+
"learning_rate": 0.0004069069069069069,
|
| 1220 |
+
"loss": 0.8942,
|
| 1221 |
+
"step": 1980
|
| 1222 |
+
},
|
| 1223 |
+
{
|
| 1224 |
+
"epoch": 2.69,
|
| 1225 |
+
"learning_rate": 0.00040615615615615617,
|
| 1226 |
+
"loss": 0.9473,
|
| 1227 |
+
"step": 1990
|
| 1228 |
+
},
|
| 1229 |
+
{
|
| 1230 |
+
"epoch": 2.7,
|
| 1231 |
+
"learning_rate": 0.00040540540540540544,
|
| 1232 |
+
"loss": 0.9547,
|
| 1233 |
+
"step": 2000
|
| 1234 |
+
},
|
| 1235 |
+
{
|
| 1236 |
+
"epoch": 2.71,
|
| 1237 |
+
"learning_rate": 0.00040465465465465466,
|
| 1238 |
+
"loss": 0.9377,
|
| 1239 |
+
"step": 2010
|
| 1240 |
+
},
|
| 1241 |
+
{
|
| 1242 |
+
"epoch": 2.73,
|
| 1243 |
+
"learning_rate": 0.00040390390390390393,
|
| 1244 |
+
"loss": 0.9604,
|
| 1245 |
+
"step": 2020
|
| 1246 |
+
},
|
| 1247 |
+
{
|
| 1248 |
+
"epoch": 2.74,
|
| 1249 |
+
"learning_rate": 0.00040315315315315314,
|
| 1250 |
+
"loss": 0.9484,
|
| 1251 |
+
"step": 2030
|
| 1252 |
+
},
|
| 1253 |
+
{
|
| 1254 |
+
"epoch": 2.75,
|
| 1255 |
+
"learning_rate": 0.0004024024024024024,
|
| 1256 |
+
"loss": 0.8361,
|
| 1257 |
+
"step": 2040
|
| 1258 |
+
},
|
| 1259 |
+
{
|
| 1260 |
+
"epoch": 2.77,
|
| 1261 |
+
"learning_rate": 0.0004016516516516517,
|
| 1262 |
+
"loss": 0.8545,
|
| 1263 |
+
"step": 2050
|
| 1264 |
+
},
|
| 1265 |
+
{
|
| 1266 |
+
"epoch": 2.78,
|
| 1267 |
+
"learning_rate": 0.0004009009009009009,
|
| 1268 |
+
"loss": 0.9045,
|
| 1269 |
+
"step": 2060
|
| 1270 |
+
},
|
| 1271 |
+
{
|
| 1272 |
+
"epoch": 2.79,
|
| 1273 |
+
"learning_rate": 0.0004001501501501502,
|
| 1274 |
+
"loss": 0.8322,
|
| 1275 |
+
"step": 2070
|
| 1276 |
+
},
|
| 1277 |
+
{
|
| 1278 |
+
"epoch": 2.81,
|
| 1279 |
+
"learning_rate": 0.0003993993993993994,
|
| 1280 |
+
"loss": 0.9181,
|
| 1281 |
+
"step": 2080
|
| 1282 |
+
},
|
| 1283 |
+
{
|
| 1284 |
+
"epoch": 2.82,
|
| 1285 |
+
"learning_rate": 0.00039864864864864866,
|
| 1286 |
+
"loss": 0.7966,
|
| 1287 |
+
"step": 2090
|
| 1288 |
+
},
|
| 1289 |
+
{
|
| 1290 |
+
"epoch": 2.83,
|
| 1291 |
+
"learning_rate": 0.00039789789789789794,
|
| 1292 |
+
"loss": 0.9089,
|
| 1293 |
+
"step": 2100
|
| 1294 |
+
},
|
| 1295 |
+
{
|
| 1296 |
+
"epoch": 2.85,
|
| 1297 |
+
"learning_rate": 0.00039714714714714715,
|
| 1298 |
+
"loss": 0.9082,
|
| 1299 |
+
"step": 2110
|
| 1300 |
+
},
|
| 1301 |
+
{
|
| 1302 |
+
"epoch": 2.86,
|
| 1303 |
+
"learning_rate": 0.0003963963963963964,
|
| 1304 |
+
"loss": 0.8908,
|
| 1305 |
+
"step": 2120
|
| 1306 |
+
},
|
| 1307 |
+
{
|
| 1308 |
+
"epoch": 2.88,
|
| 1309 |
+
"learning_rate": 0.00039564564564564564,
|
| 1310 |
+
"loss": 0.8805,
|
| 1311 |
+
"step": 2130
|
| 1312 |
+
},
|
| 1313 |
+
{
|
| 1314 |
+
"epoch": 2.89,
|
| 1315 |
+
"learning_rate": 0.00039489489489489486,
|
| 1316 |
+
"loss": 0.9338,
|
| 1317 |
+
"step": 2140
|
| 1318 |
+
},
|
| 1319 |
+
{
|
| 1320 |
+
"epoch": 2.9,
|
| 1321 |
+
"learning_rate": 0.0003941441441441442,
|
| 1322 |
+
"loss": 0.9174,
|
| 1323 |
+
"step": 2150
|
| 1324 |
+
},
|
| 1325 |
+
{
|
| 1326 |
+
"epoch": 2.92,
|
| 1327 |
+
"learning_rate": 0.0003933933933933934,
|
| 1328 |
+
"loss": 0.902,
|
| 1329 |
+
"step": 2160
|
| 1330 |
+
},
|
| 1331 |
+
{
|
| 1332 |
+
"epoch": 2.93,
|
| 1333 |
+
"learning_rate": 0.00039264264264264267,
|
| 1334 |
+
"loss": 0.8365,
|
| 1335 |
+
"step": 2170
|
| 1336 |
+
},
|
| 1337 |
+
{
|
| 1338 |
+
"epoch": 2.94,
|
| 1339 |
+
"learning_rate": 0.0003918918918918919,
|
| 1340 |
+
"loss": 0.7663,
|
| 1341 |
+
"step": 2180
|
| 1342 |
+
},
|
| 1343 |
+
{
|
| 1344 |
+
"epoch": 2.96,
|
| 1345 |
+
"learning_rate": 0.0003911411411411411,
|
| 1346 |
+
"loss": 0.9336,
|
| 1347 |
+
"step": 2190
|
| 1348 |
+
},
|
| 1349 |
+
{
|
| 1350 |
+
"epoch": 2.97,
|
| 1351 |
+
"learning_rate": 0.0003903903903903904,
|
| 1352 |
+
"loss": 0.8055,
|
| 1353 |
+
"step": 2200
|
| 1354 |
+
},
|
| 1355 |
+
{
|
| 1356 |
+
"epoch": 2.98,
|
| 1357 |
+
"learning_rate": 0.00038963963963963965,
|
| 1358 |
+
"loss": 0.8888,
|
| 1359 |
+
"step": 2210
|
| 1360 |
+
},
|
| 1361 |
+
{
|
| 1362 |
+
"epoch": 3.0,
|
| 1363 |
+
"learning_rate": 0.0003888888888888889,
|
| 1364 |
+
"loss": 0.9069,
|
| 1365 |
+
"step": 2220
|
| 1366 |
+
},
|
| 1367 |
+
{
|
| 1368 |
+
"epoch": 3.0,
|
| 1369 |
+
"eval_accuracy": 0.8292658531706342,
|
| 1370 |
+
"eval_f1": 0.8288337486077575,
|
| 1371 |
+
"eval_loss": 0.5952173471450806,
|
| 1372 |
+
"eval_precision": 0.842670072902701,
|
| 1373 |
+
"eval_recall": 0.8282744379906413,
|
| 1374 |
+
"eval_runtime": 192.1316,
|
| 1375 |
+
"eval_samples_per_second": 52.037,
|
| 1376 |
+
"eval_steps_per_second": 1.931,
|
| 1377 |
+
"step": 2222
|
| 1378 |
+
},
|
| 1379 |
+
{
|
| 1380 |
+
"epoch": 3.01,
|
| 1381 |
+
"learning_rate": 0.00038813813813813814,
|
| 1382 |
+
"loss": 0.7958,
|
| 1383 |
+
"step": 2230
|
| 1384 |
+
},
|
| 1385 |
+
{
|
| 1386 |
+
"epoch": 3.02,
|
| 1387 |
+
"learning_rate": 0.00038738738738738736,
|
| 1388 |
+
"loss": 0.7455,
|
| 1389 |
+
"step": 2240
|
| 1390 |
+
},
|
| 1391 |
+
{
|
| 1392 |
+
"epoch": 3.04,
|
| 1393 |
+
"learning_rate": 0.0003866366366366366,
|
| 1394 |
+
"loss": 0.7151,
|
| 1395 |
+
"step": 2250
|
| 1396 |
+
},
|
| 1397 |
+
{
|
| 1398 |
+
"epoch": 3.05,
|
| 1399 |
+
"learning_rate": 0.0003858858858858859,
|
| 1400 |
+
"loss": 0.7621,
|
| 1401 |
+
"step": 2260
|
| 1402 |
+
},
|
| 1403 |
+
{
|
| 1404 |
+
"epoch": 3.06,
|
| 1405 |
+
"learning_rate": 0.00038513513513513517,
|
| 1406 |
+
"loss": 0.8418,
|
| 1407 |
+
"step": 2270
|
| 1408 |
+
},
|
| 1409 |
+
{
|
| 1410 |
+
"epoch": 3.08,
|
| 1411 |
+
"learning_rate": 0.0003843843843843844,
|
| 1412 |
+
"loss": 0.8065,
|
| 1413 |
+
"step": 2280
|
| 1414 |
+
},
|
| 1415 |
+
{
|
| 1416 |
+
"epoch": 3.09,
|
| 1417 |
+
"learning_rate": 0.0003836336336336336,
|
| 1418 |
+
"loss": 0.7361,
|
| 1419 |
+
"step": 2290
|
| 1420 |
+
},
|
| 1421 |
+
{
|
| 1422 |
+
"epoch": 3.1,
|
| 1423 |
+
"learning_rate": 0.0003828828828828829,
|
| 1424 |
+
"loss": 0.7774,
|
| 1425 |
+
"step": 2300
|
| 1426 |
+
},
|
| 1427 |
+
{
|
| 1428 |
+
"epoch": 3.12,
|
| 1429 |
+
"learning_rate": 0.00038213213213213215,
|
| 1430 |
+
"loss": 0.7262,
|
| 1431 |
+
"step": 2310
|
| 1432 |
+
},
|
| 1433 |
+
{
|
| 1434 |
+
"epoch": 3.13,
|
| 1435 |
+
"learning_rate": 0.0003813813813813814,
|
| 1436 |
+
"loss": 0.7989,
|
| 1437 |
+
"step": 2320
|
| 1438 |
+
},
|
| 1439 |
+
{
|
| 1440 |
+
"epoch": 3.15,
|
| 1441 |
+
"learning_rate": 0.00038063063063063064,
|
| 1442 |
+
"loss": 0.8062,
|
| 1443 |
+
"step": 2330
|
| 1444 |
+
},
|
| 1445 |
+
{
|
| 1446 |
+
"epoch": 3.16,
|
| 1447 |
+
"learning_rate": 0.00037987987987987985,
|
| 1448 |
+
"loss": 0.7957,
|
| 1449 |
+
"step": 2340
|
| 1450 |
+
},
|
| 1451 |
+
{
|
| 1452 |
+
"epoch": 3.17,
|
| 1453 |
+
"learning_rate": 0.0003791291291291291,
|
| 1454 |
+
"loss": 0.7507,
|
| 1455 |
+
"step": 2350
|
| 1456 |
+
},
|
| 1457 |
+
{
|
| 1458 |
+
"epoch": 3.19,
|
| 1459 |
+
"learning_rate": 0.0003783783783783784,
|
| 1460 |
+
"loss": 0.7881,
|
| 1461 |
+
"step": 2360
|
| 1462 |
+
},
|
| 1463 |
+
{
|
| 1464 |
+
"epoch": 3.2,
|
| 1465 |
+
"learning_rate": 0.00037762762762762767,
|
| 1466 |
+
"loss": 0.7048,
|
| 1467 |
+
"step": 2370
|
| 1468 |
+
},
|
| 1469 |
+
{
|
| 1470 |
+
"epoch": 3.21,
|
| 1471 |
+
"learning_rate": 0.0003768768768768769,
|
| 1472 |
+
"loss": 0.8162,
|
| 1473 |
+
"step": 2380
|
| 1474 |
+
},
|
| 1475 |
+
{
|
| 1476 |
+
"epoch": 3.23,
|
| 1477 |
+
"learning_rate": 0.0003761261261261261,
|
| 1478 |
+
"loss": 0.7446,
|
| 1479 |
+
"step": 2390
|
| 1480 |
+
},
|
| 1481 |
+
{
|
| 1482 |
+
"epoch": 3.24,
|
| 1483 |
+
"learning_rate": 0.00037537537537537537,
|
| 1484 |
+
"loss": 0.7904,
|
| 1485 |
+
"step": 2400
|
| 1486 |
+
},
|
| 1487 |
+
{
|
| 1488 |
+
"epoch": 3.25,
|
| 1489 |
+
"learning_rate": 0.00037462462462462464,
|
| 1490 |
+
"loss": 0.7939,
|
| 1491 |
+
"step": 2410
|
| 1492 |
+
},
|
| 1493 |
+
{
|
| 1494 |
+
"epoch": 3.27,
|
| 1495 |
+
"learning_rate": 0.0003738738738738739,
|
| 1496 |
+
"loss": 0.6914,
|
| 1497 |
+
"step": 2420
|
| 1498 |
+
},
|
| 1499 |
+
{
|
| 1500 |
+
"epoch": 3.28,
|
| 1501 |
+
"learning_rate": 0.00037312312312312313,
|
| 1502 |
+
"loss": 0.8431,
|
| 1503 |
+
"step": 2430
|
| 1504 |
+
},
|
| 1505 |
+
{
|
| 1506 |
+
"epoch": 3.29,
|
| 1507 |
+
"learning_rate": 0.00037237237237237235,
|
| 1508 |
+
"loss": 0.775,
|
| 1509 |
+
"step": 2440
|
| 1510 |
+
},
|
| 1511 |
+
{
|
| 1512 |
+
"epoch": 3.31,
|
| 1513 |
+
"learning_rate": 0.0003716216216216216,
|
| 1514 |
+
"loss": 0.7682,
|
| 1515 |
+
"step": 2450
|
| 1516 |
+
},
|
| 1517 |
+
{
|
| 1518 |
+
"epoch": 3.32,
|
| 1519 |
+
"learning_rate": 0.0003708708708708709,
|
| 1520 |
+
"loss": 0.8463,
|
| 1521 |
+
"step": 2460
|
| 1522 |
+
},
|
| 1523 |
+
{
|
| 1524 |
+
"epoch": 3.33,
|
| 1525 |
+
"learning_rate": 0.00037012012012012016,
|
| 1526 |
+
"loss": 0.7268,
|
| 1527 |
+
"step": 2470
|
| 1528 |
+
},
|
| 1529 |
+
{
|
| 1530 |
+
"epoch": 3.35,
|
| 1531 |
+
"learning_rate": 0.0003693693693693694,
|
| 1532 |
+
"loss": 0.7556,
|
| 1533 |
+
"step": 2480
|
| 1534 |
+
},
|
| 1535 |
+
{
|
| 1536 |
+
"epoch": 3.36,
|
| 1537 |
+
"learning_rate": 0.0003686186186186186,
|
| 1538 |
+
"loss": 0.7808,
|
| 1539 |
+
"step": 2490
|
| 1540 |
+
},
|
| 1541 |
+
{
|
| 1542 |
+
"epoch": 3.37,
|
| 1543 |
+
"learning_rate": 0.00036786786786786787,
|
| 1544 |
+
"loss": 0.7615,
|
| 1545 |
+
"step": 2500
|
| 1546 |
+
},
|
| 1547 |
+
{
|
| 1548 |
+
"epoch": 3.39,
|
| 1549 |
+
"learning_rate": 0.00036711711711711714,
|
| 1550 |
+
"loss": 0.798,
|
| 1551 |
+
"step": 2510
|
| 1552 |
+
},
|
| 1553 |
+
{
|
| 1554 |
+
"epoch": 3.4,
|
| 1555 |
+
"learning_rate": 0.0003663663663663664,
|
| 1556 |
+
"loss": 0.7917,
|
| 1557 |
+
"step": 2520
|
| 1558 |
+
},
|
| 1559 |
+
{
|
| 1560 |
+
"epoch": 3.42,
|
| 1561 |
+
"learning_rate": 0.00036561561561561563,
|
| 1562 |
+
"loss": 0.75,
|
| 1563 |
+
"step": 2530
|
| 1564 |
+
},
|
| 1565 |
+
{
|
| 1566 |
+
"epoch": 3.43,
|
| 1567 |
+
"learning_rate": 0.00036486486486486485,
|
| 1568 |
+
"loss": 0.6937,
|
| 1569 |
+
"step": 2540
|
| 1570 |
+
},
|
| 1571 |
+
{
|
| 1572 |
+
"epoch": 3.44,
|
| 1573 |
+
"learning_rate": 0.0003641141141141141,
|
| 1574 |
+
"loss": 0.7988,
|
| 1575 |
+
"step": 2550
|
| 1576 |
+
},
|
| 1577 |
+
{
|
| 1578 |
+
"epoch": 3.46,
|
| 1579 |
+
"learning_rate": 0.0003633633633633634,
|
| 1580 |
+
"loss": 0.7315,
|
| 1581 |
+
"step": 2560
|
| 1582 |
+
},
|
| 1583 |
+
{
|
| 1584 |
+
"epoch": 3.47,
|
| 1585 |
+
"learning_rate": 0.00036261261261261266,
|
| 1586 |
+
"loss": 0.7546,
|
| 1587 |
+
"step": 2570
|
| 1588 |
+
},
|
| 1589 |
+
{
|
| 1590 |
+
"epoch": 3.48,
|
| 1591 |
+
"learning_rate": 0.0003618618618618619,
|
| 1592 |
+
"loss": 0.7216,
|
| 1593 |
+
"step": 2580
|
| 1594 |
+
},
|
| 1595 |
+
{
|
| 1596 |
+
"epoch": 3.5,
|
| 1597 |
+
"learning_rate": 0.0003611111111111111,
|
| 1598 |
+
"loss": 0.7669,
|
| 1599 |
+
"step": 2590
|
| 1600 |
+
},
|
| 1601 |
+
{
|
| 1602 |
+
"epoch": 3.51,
|
| 1603 |
+
"learning_rate": 0.00036036036036036037,
|
| 1604 |
+
"loss": 0.7127,
|
| 1605 |
+
"step": 2600
|
| 1606 |
+
},
|
| 1607 |
+
{
|
| 1608 |
+
"epoch": 3.52,
|
| 1609 |
+
"learning_rate": 0.00035960960960960964,
|
| 1610 |
+
"loss": 0.7755,
|
| 1611 |
+
"step": 2610
|
| 1612 |
+
},
|
| 1613 |
+
{
|
| 1614 |
+
"epoch": 3.54,
|
| 1615 |
+
"learning_rate": 0.0003588588588588589,
|
| 1616 |
+
"loss": 0.8252,
|
| 1617 |
+
"step": 2620
|
| 1618 |
+
},
|
| 1619 |
+
{
|
| 1620 |
+
"epoch": 3.55,
|
| 1621 |
+
"learning_rate": 0.0003581081081081081,
|
| 1622 |
+
"loss": 0.7839,
|
| 1623 |
+
"step": 2630
|
| 1624 |
+
},
|
| 1625 |
+
{
|
| 1626 |
+
"epoch": 3.56,
|
| 1627 |
+
"learning_rate": 0.00035735735735735734,
|
| 1628 |
+
"loss": 0.7111,
|
| 1629 |
+
"step": 2640
|
| 1630 |
+
},
|
| 1631 |
+
{
|
| 1632 |
+
"epoch": 3.58,
|
| 1633 |
+
"learning_rate": 0.0003566066066066066,
|
| 1634 |
+
"loss": 0.6958,
|
| 1635 |
+
"step": 2650
|
| 1636 |
+
},
|
| 1637 |
+
{
|
| 1638 |
+
"epoch": 3.59,
|
| 1639 |
+
"learning_rate": 0.00035585585585585583,
|
| 1640 |
+
"loss": 0.6712,
|
| 1641 |
+
"step": 2660
|
| 1642 |
+
},
|
| 1643 |
+
{
|
| 1644 |
+
"epoch": 3.6,
|
| 1645 |
+
"learning_rate": 0.00035510510510510516,
|
| 1646 |
+
"loss": 0.6692,
|
| 1647 |
+
"step": 2670
|
| 1648 |
+
},
|
| 1649 |
+
{
|
| 1650 |
+
"epoch": 3.62,
|
| 1651 |
+
"learning_rate": 0.0003543543543543544,
|
| 1652 |
+
"loss": 0.7866,
|
| 1653 |
+
"step": 2680
|
| 1654 |
+
},
|
| 1655 |
+
{
|
| 1656 |
+
"epoch": 3.63,
|
| 1657 |
+
"learning_rate": 0.0003536036036036036,
|
| 1658 |
+
"loss": 0.7061,
|
| 1659 |
+
"step": 2690
|
| 1660 |
+
},
|
| 1661 |
+
{
|
| 1662 |
+
"epoch": 3.64,
|
| 1663 |
+
"learning_rate": 0.00035285285285285286,
|
| 1664 |
+
"loss": 0.7742,
|
| 1665 |
+
"step": 2700
|
| 1666 |
+
},
|
| 1667 |
+
{
|
| 1668 |
+
"epoch": 3.66,
|
| 1669 |
+
"learning_rate": 0.0003521021021021021,
|
| 1670 |
+
"loss": 0.6905,
|
| 1671 |
+
"step": 2710
|
| 1672 |
+
},
|
| 1673 |
+
{
|
| 1674 |
+
"epoch": 3.67,
|
| 1675 |
+
"learning_rate": 0.00035135135135135135,
|
| 1676 |
+
"loss": 0.7464,
|
| 1677 |
+
"step": 2720
|
| 1678 |
+
},
|
| 1679 |
+
{
|
| 1680 |
+
"epoch": 3.69,
|
| 1681 |
+
"learning_rate": 0.0003506006006006006,
|
| 1682 |
+
"loss": 0.7855,
|
| 1683 |
+
"step": 2730
|
| 1684 |
+
},
|
| 1685 |
+
{
|
| 1686 |
+
"epoch": 3.7,
|
| 1687 |
+
"learning_rate": 0.00034984984984984984,
|
| 1688 |
+
"loss": 0.7871,
|
| 1689 |
+
"step": 2740
|
| 1690 |
+
},
|
| 1691 |
+
{
|
| 1692 |
+
"epoch": 3.71,
|
| 1693 |
+
"learning_rate": 0.0003490990990990991,
|
| 1694 |
+
"loss": 0.6782,
|
| 1695 |
+
"step": 2750
|
| 1696 |
+
},
|
| 1697 |
+
{
|
| 1698 |
+
"epoch": 3.73,
|
| 1699 |
+
"learning_rate": 0.00034834834834834833,
|
| 1700 |
+
"loss": 0.836,
|
| 1701 |
+
"step": 2760
|
| 1702 |
+
},
|
| 1703 |
+
{
|
| 1704 |
+
"epoch": 3.74,
|
| 1705 |
+
"learning_rate": 0.0003475975975975976,
|
| 1706 |
+
"loss": 0.6404,
|
| 1707 |
+
"step": 2770
|
| 1708 |
+
},
|
| 1709 |
+
{
|
| 1710 |
+
"epoch": 3.75,
|
| 1711 |
+
"learning_rate": 0.00034684684684684687,
|
| 1712 |
+
"loss": 0.8315,
|
| 1713 |
+
"step": 2780
|
| 1714 |
+
},
|
| 1715 |
+
{
|
| 1716 |
+
"epoch": 3.77,
|
| 1717 |
+
"learning_rate": 0.0003460960960960961,
|
| 1718 |
+
"loss": 0.6882,
|
| 1719 |
+
"step": 2790
|
| 1720 |
+
},
|
| 1721 |
+
{
|
| 1722 |
+
"epoch": 3.78,
|
| 1723 |
+
"learning_rate": 0.00034534534534534536,
|
| 1724 |
+
"loss": 0.7388,
|
| 1725 |
+
"step": 2800
|
| 1726 |
+
},
|
| 1727 |
+
{
|
| 1728 |
+
"epoch": 3.79,
|
| 1729 |
+
"learning_rate": 0.0003445945945945946,
|
| 1730 |
+
"loss": 0.8343,
|
| 1731 |
+
"step": 2810
|
| 1732 |
+
},
|
| 1733 |
+
{
|
| 1734 |
+
"epoch": 3.81,
|
| 1735 |
+
"learning_rate": 0.00034384384384384385,
|
| 1736 |
+
"loss": 0.7021,
|
| 1737 |
+
"step": 2820
|
| 1738 |
+
},
|
| 1739 |
+
{
|
| 1740 |
+
"epoch": 3.82,
|
| 1741 |
+
"learning_rate": 0.0003430930930930931,
|
| 1742 |
+
"loss": 0.7739,
|
| 1743 |
+
"step": 2830
|
| 1744 |
+
},
|
| 1745 |
+
{
|
| 1746 |
+
"epoch": 3.83,
|
| 1747 |
+
"learning_rate": 0.00034234234234234234,
|
| 1748 |
+
"loss": 0.7021,
|
| 1749 |
+
"step": 2840
|
| 1750 |
+
},
|
| 1751 |
+
{
|
| 1752 |
+
"epoch": 3.85,
|
| 1753 |
+
"learning_rate": 0.0003415915915915916,
|
| 1754 |
+
"loss": 0.8136,
|
| 1755 |
+
"step": 2850
|
| 1756 |
+
},
|
| 1757 |
+
{
|
| 1758 |
+
"epoch": 3.86,
|
| 1759 |
+
"learning_rate": 0.0003408408408408408,
|
| 1760 |
+
"loss": 0.7634,
|
| 1761 |
+
"step": 2860
|
| 1762 |
+
},
|
| 1763 |
+
{
|
| 1764 |
+
"epoch": 3.87,
|
| 1765 |
+
"learning_rate": 0.0003400900900900901,
|
| 1766 |
+
"loss": 0.7506,
|
| 1767 |
+
"step": 2870
|
| 1768 |
+
},
|
| 1769 |
+
{
|
| 1770 |
+
"epoch": 3.89,
|
| 1771 |
+
"learning_rate": 0.00033933933933933937,
|
| 1772 |
+
"loss": 0.7432,
|
| 1773 |
+
"step": 2880
|
| 1774 |
+
},
|
| 1775 |
+
{
|
| 1776 |
+
"epoch": 3.9,
|
| 1777 |
+
"learning_rate": 0.0003385885885885886,
|
| 1778 |
+
"loss": 0.7378,
|
| 1779 |
+
"step": 2890
|
| 1780 |
+
},
|
| 1781 |
+
{
|
| 1782 |
+
"epoch": 3.91,
|
| 1783 |
+
"learning_rate": 0.00033783783783783786,
|
| 1784 |
+
"loss": 0.6994,
|
| 1785 |
+
"step": 2900
|
| 1786 |
+
},
|
| 1787 |
+
{
|
| 1788 |
+
"epoch": 3.93,
|
| 1789 |
+
"learning_rate": 0.0003370870870870871,
|
| 1790 |
+
"loss": 0.7371,
|
| 1791 |
+
"step": 2910
|
| 1792 |
+
},
|
| 1793 |
+
{
|
| 1794 |
+
"epoch": 3.94,
|
| 1795 |
+
"learning_rate": 0.00033633633633633635,
|
| 1796 |
+
"loss": 0.7055,
|
| 1797 |
+
"step": 2920
|
| 1798 |
+
},
|
| 1799 |
+
{
|
| 1800 |
+
"epoch": 3.96,
|
| 1801 |
+
"learning_rate": 0.0003355855855855856,
|
| 1802 |
+
"loss": 0.703,
|
| 1803 |
+
"step": 2930
|
| 1804 |
+
},
|
| 1805 |
+
{
|
| 1806 |
+
"epoch": 3.97,
|
| 1807 |
+
"learning_rate": 0.00033483483483483483,
|
| 1808 |
+
"loss": 0.7472,
|
| 1809 |
+
"step": 2940
|
| 1810 |
+
},
|
| 1811 |
+
{
|
| 1812 |
+
"epoch": 3.98,
|
| 1813 |
+
"learning_rate": 0.0003340840840840841,
|
| 1814 |
+
"loss": 0.8109,
|
| 1815 |
+
"step": 2950
|
| 1816 |
+
},
|
| 1817 |
+
{
|
| 1818 |
+
"epoch": 4.0,
|
| 1819 |
+
"learning_rate": 0.0003333333333333333,
|
| 1820 |
+
"loss": 0.7519,
|
| 1821 |
+
"step": 2960
|
| 1822 |
+
},
|
| 1823 |
+
{
|
| 1824 |
+
"epoch": 4.0,
|
| 1825 |
+
"eval_accuracy": 0.8491698339667934,
|
| 1826 |
+
"eval_f1": 0.849423792771648,
|
| 1827 |
+
"eval_loss": 0.5195841789245605,
|
| 1828 |
+
"eval_precision": 0.8611084598890817,
|
| 1829 |
+
"eval_recall": 0.8482889408623805,
|
| 1830 |
+
"eval_runtime": 190.693,
|
| 1831 |
+
"eval_samples_per_second": 52.43,
|
| 1832 |
+
"eval_steps_per_second": 1.946,
|
| 1833 |
+
"step": 2963
|
| 1834 |
+
},
|
| 1835 |
+
{
|
| 1836 |
+
"epoch": 4.01,
|
| 1837 |
+
"learning_rate": 0.0003325825825825826,
|
| 1838 |
+
"loss": 0.7232,
|
| 1839 |
+
"step": 2970
|
| 1840 |
+
},
|
| 1841 |
+
{
|
| 1842 |
+
"epoch": 4.02,
|
| 1843 |
+
"learning_rate": 0.00033183183183183186,
|
| 1844 |
+
"loss": 0.5435,
|
| 1845 |
+
"step": 2980
|
| 1846 |
+
},
|
| 1847 |
+
{
|
| 1848 |
+
"epoch": 4.04,
|
| 1849 |
+
"learning_rate": 0.0003310810810810811,
|
| 1850 |
+
"loss": 0.6165,
|
| 1851 |
+
"step": 2990
|
| 1852 |
+
},
|
| 1853 |
+
{
|
| 1854 |
+
"epoch": 4.05,
|
| 1855 |
+
"learning_rate": 0.00033033033033033035,
|
| 1856 |
+
"loss": 0.6642,
|
| 1857 |
+
"step": 3000
|
| 1858 |
+
},
|
| 1859 |
+
{
|
| 1860 |
+
"epoch": 4.06,
|
| 1861 |
+
"learning_rate": 0.00032957957957957957,
|
| 1862 |
+
"loss": 0.6566,
|
| 1863 |
+
"step": 3010
|
| 1864 |
+
},
|
| 1865 |
+
{
|
| 1866 |
+
"epoch": 4.08,
|
| 1867 |
+
"learning_rate": 0.00032882882882882884,
|
| 1868 |
+
"loss": 0.6535,
|
| 1869 |
+
"step": 3020
|
| 1870 |
+
},
|
| 1871 |
+
{
|
| 1872 |
+
"epoch": 4.09,
|
| 1873 |
+
"learning_rate": 0.0003280780780780781,
|
| 1874 |
+
"loss": 0.6825,
|
| 1875 |
+
"step": 3030
|
| 1876 |
+
},
|
| 1877 |
+
{
|
| 1878 |
+
"epoch": 4.1,
|
| 1879 |
+
"learning_rate": 0.00032732732732732733,
|
| 1880 |
+
"loss": 0.5755,
|
| 1881 |
+
"step": 3040
|
| 1882 |
+
},
|
| 1883 |
+
{
|
| 1884 |
+
"epoch": 4.12,
|
| 1885 |
+
"learning_rate": 0.0003265765765765766,
|
| 1886 |
+
"loss": 0.6186,
|
| 1887 |
+
"step": 3050
|
| 1888 |
+
},
|
| 1889 |
+
{
|
| 1890 |
+
"epoch": 4.13,
|
| 1891 |
+
"learning_rate": 0.0003258258258258258,
|
| 1892 |
+
"loss": 0.6469,
|
| 1893 |
+
"step": 3060
|
| 1894 |
+
},
|
| 1895 |
+
{
|
| 1896 |
+
"epoch": 4.14,
|
| 1897 |
+
"learning_rate": 0.00032507507507507504,
|
| 1898 |
+
"loss": 0.6201,
|
| 1899 |
+
"step": 3070
|
| 1900 |
+
},
|
| 1901 |
+
{
|
| 1902 |
+
"epoch": 4.16,
|
| 1903 |
+
"learning_rate": 0.00032432432432432436,
|
| 1904 |
+
"loss": 0.6528,
|
| 1905 |
+
"step": 3080
|
| 1906 |
+
},
|
| 1907 |
+
{
|
| 1908 |
+
"epoch": 4.17,
|
| 1909 |
+
"learning_rate": 0.0003235735735735736,
|
| 1910 |
+
"loss": 0.7014,
|
| 1911 |
+
"step": 3090
|
| 1912 |
+
},
|
| 1913 |
+
{
|
| 1914 |
+
"epoch": 4.18,
|
| 1915 |
+
"learning_rate": 0.00032282282282282285,
|
| 1916 |
+
"loss": 0.6835,
|
| 1917 |
+
"step": 3100
|
| 1918 |
+
},
|
| 1919 |
+
{
|
| 1920 |
+
"epoch": 4.2,
|
| 1921 |
+
"learning_rate": 0.00032207207207207207,
|
| 1922 |
+
"loss": 0.6827,
|
| 1923 |
+
"step": 3110
|
| 1924 |
+
},
|
| 1925 |
+
{
|
| 1926 |
+
"epoch": 4.21,
|
| 1927 |
+
"learning_rate": 0.0003213213213213213,
|
| 1928 |
+
"loss": 0.7095,
|
| 1929 |
+
"step": 3120
|
| 1930 |
+
},
|
| 1931 |
+
{
|
| 1932 |
+
"epoch": 4.23,
|
| 1933 |
+
"learning_rate": 0.0003205705705705706,
|
| 1934 |
+
"loss": 0.6773,
|
| 1935 |
+
"step": 3130
|
| 1936 |
+
},
|
| 1937 |
+
{
|
| 1938 |
+
"epoch": 4.24,
|
| 1939 |
+
"learning_rate": 0.00031981981981981983,
|
| 1940 |
+
"loss": 0.6967,
|
| 1941 |
+
"step": 3140
|
| 1942 |
+
},
|
| 1943 |
+
{
|
| 1944 |
+
"epoch": 4.25,
|
| 1945 |
+
"learning_rate": 0.0003190690690690691,
|
| 1946 |
+
"loss": 0.6751,
|
| 1947 |
+
"step": 3150
|
| 1948 |
+
},
|
| 1949 |
+
{
|
| 1950 |
+
"epoch": 4.27,
|
| 1951 |
+
"learning_rate": 0.0003183183183183183,
|
| 1952 |
+
"loss": 0.6711,
|
| 1953 |
+
"step": 3160
|
| 1954 |
+
},
|
| 1955 |
+
{
|
| 1956 |
+
"epoch": 4.28,
|
| 1957 |
+
"learning_rate": 0.00031756756756756753,
|
| 1958 |
+
"loss": 0.5826,
|
| 1959 |
+
"step": 3170
|
| 1960 |
+
},
|
| 1961 |
+
{
|
| 1962 |
+
"epoch": 4.29,
|
| 1963 |
+
"learning_rate": 0.00031681681681681686,
|
| 1964 |
+
"loss": 0.6481,
|
| 1965 |
+
"step": 3180
|
| 1966 |
+
},
|
| 1967 |
+
{
|
| 1968 |
+
"epoch": 4.31,
|
| 1969 |
+
"learning_rate": 0.0003160660660660661,
|
| 1970 |
+
"loss": 0.6282,
|
| 1971 |
+
"step": 3190
|
| 1972 |
+
},
|
| 1973 |
+
{
|
| 1974 |
+
"epoch": 4.32,
|
| 1975 |
+
"learning_rate": 0.00031531531531531535,
|
| 1976 |
+
"loss": 0.6441,
|
| 1977 |
+
"step": 3200
|
| 1978 |
+
},
|
| 1979 |
+
{
|
| 1980 |
+
"epoch": 4.33,
|
| 1981 |
+
"learning_rate": 0.00031456456456456456,
|
| 1982 |
+
"loss": 0.643,
|
| 1983 |
+
"step": 3210
|
| 1984 |
+
},
|
| 1985 |
+
{
|
| 1986 |
+
"epoch": 4.35,
|
| 1987 |
+
"learning_rate": 0.0003138138138138138,
|
| 1988 |
+
"loss": 0.6681,
|
| 1989 |
+
"step": 3220
|
| 1990 |
+
},
|
| 1991 |
+
{
|
| 1992 |
+
"epoch": 4.36,
|
| 1993 |
+
"learning_rate": 0.0003130630630630631,
|
| 1994 |
+
"loss": 0.642,
|
| 1995 |
+
"step": 3230
|
| 1996 |
+
},
|
| 1997 |
+
{
|
| 1998 |
+
"epoch": 4.37,
|
| 1999 |
+
"learning_rate": 0.0003123123123123123,
|
| 2000 |
+
"loss": 0.6114,
|
| 2001 |
+
"step": 3240
|
| 2002 |
+
},
|
| 2003 |
+
{
|
| 2004 |
+
"epoch": 4.39,
|
| 2005 |
+
"learning_rate": 0.00031156156156156154,
|
| 2006 |
+
"loss": 0.6465,
|
| 2007 |
+
"step": 3250
|
| 2008 |
+
},
|
| 2009 |
+
{
|
| 2010 |
+
"epoch": 4.4,
|
| 2011 |
+
"learning_rate": 0.0003108108108108108,
|
| 2012 |
+
"loss": 0.7339,
|
| 2013 |
+
"step": 3260
|
| 2014 |
+
},
|
| 2015 |
+
{
|
| 2016 |
+
"epoch": 4.41,
|
| 2017 |
+
"learning_rate": 0.00031006006006006003,
|
| 2018 |
+
"loss": 0.6085,
|
| 2019 |
+
"step": 3270
|
| 2020 |
+
},
|
| 2021 |
+
{
|
| 2022 |
+
"epoch": 4.43,
|
| 2023 |
+
"learning_rate": 0.00030930930930930936,
|
| 2024 |
+
"loss": 0.6396,
|
| 2025 |
+
"step": 3280
|
| 2026 |
+
},
|
| 2027 |
+
{
|
| 2028 |
+
"epoch": 4.44,
|
| 2029 |
+
"learning_rate": 0.00030855855855855857,
|
| 2030 |
+
"loss": 0.6864,
|
| 2031 |
+
"step": 3290
|
| 2032 |
+
},
|
| 2033 |
+
{
|
| 2034 |
+
"epoch": 4.45,
|
| 2035 |
+
"learning_rate": 0.0003078078078078078,
|
| 2036 |
+
"loss": 0.6241,
|
| 2037 |
+
"step": 3300
|
| 2038 |
+
},
|
| 2039 |
+
{
|
| 2040 |
+
"epoch": 4.47,
|
| 2041 |
+
"learning_rate": 0.00030705705705705706,
|
| 2042 |
+
"loss": 0.634,
|
| 2043 |
+
"step": 3310
|
| 2044 |
+
},
|
| 2045 |
+
{
|
| 2046 |
+
"epoch": 4.48,
|
| 2047 |
+
"learning_rate": 0.0003063063063063063,
|
| 2048 |
+
"loss": 0.6854,
|
| 2049 |
+
"step": 3320
|
| 2050 |
+
},
|
| 2051 |
+
{
|
| 2052 |
+
"epoch": 4.5,
|
| 2053 |
+
"learning_rate": 0.0003055555555555556,
|
| 2054 |
+
"loss": 0.5872,
|
| 2055 |
+
"step": 3330
|
| 2056 |
+
},
|
| 2057 |
+
{
|
| 2058 |
+
"epoch": 4.51,
|
| 2059 |
+
"learning_rate": 0.0003048048048048048,
|
| 2060 |
+
"loss": 0.5485,
|
| 2061 |
+
"step": 3340
|
| 2062 |
+
},
|
| 2063 |
+
{
|
| 2064 |
+
"epoch": 4.52,
|
| 2065 |
+
"learning_rate": 0.00030405405405405404,
|
| 2066 |
+
"loss": 0.6161,
|
| 2067 |
+
"step": 3350
|
| 2068 |
+
},
|
| 2069 |
+
{
|
| 2070 |
+
"epoch": 4.54,
|
| 2071 |
+
"learning_rate": 0.0003033033033033033,
|
| 2072 |
+
"loss": 0.6059,
|
| 2073 |
+
"step": 3360
|
| 2074 |
+
},
|
| 2075 |
+
{
|
| 2076 |
+
"epoch": 4.55,
|
| 2077 |
+
"learning_rate": 0.0003025525525525525,
|
| 2078 |
+
"loss": 0.6244,
|
| 2079 |
+
"step": 3370
|
| 2080 |
+
},
|
| 2081 |
+
{
|
| 2082 |
+
"epoch": 4.56,
|
| 2083 |
+
"learning_rate": 0.00030180180180180185,
|
| 2084 |
+
"loss": 0.6937,
|
| 2085 |
+
"step": 3380
|
| 2086 |
+
},
|
| 2087 |
+
{
|
| 2088 |
+
"epoch": 4.58,
|
| 2089 |
+
"learning_rate": 0.00030105105105105107,
|
| 2090 |
+
"loss": 0.6506,
|
| 2091 |
+
"step": 3390
|
| 2092 |
+
},
|
| 2093 |
+
{
|
| 2094 |
+
"epoch": 4.59,
|
| 2095 |
+
"learning_rate": 0.0003003003003003003,
|
| 2096 |
+
"loss": 0.6477,
|
| 2097 |
+
"step": 3400
|
| 2098 |
+
},
|
| 2099 |
+
{
|
| 2100 |
+
"epoch": 4.6,
|
| 2101 |
+
"learning_rate": 0.00029954954954954956,
|
| 2102 |
+
"loss": 0.6439,
|
| 2103 |
+
"step": 3410
|
| 2104 |
+
},
|
| 2105 |
+
{
|
| 2106 |
+
"epoch": 4.62,
|
| 2107 |
+
"learning_rate": 0.0002987987987987988,
|
| 2108 |
+
"loss": 0.6016,
|
| 2109 |
+
"step": 3420
|
| 2110 |
+
},
|
| 2111 |
+
{
|
| 2112 |
+
"epoch": 4.63,
|
| 2113 |
+
"learning_rate": 0.0002980480480480481,
|
| 2114 |
+
"loss": 0.5656,
|
| 2115 |
+
"step": 3430
|
| 2116 |
+
},
|
| 2117 |
+
{
|
| 2118 |
+
"epoch": 4.64,
|
| 2119 |
+
"learning_rate": 0.0002972972972972973,
|
| 2120 |
+
"loss": 0.6457,
|
| 2121 |
+
"step": 3440
|
| 2122 |
+
},
|
| 2123 |
+
{
|
| 2124 |
+
"epoch": 4.66,
|
| 2125 |
+
"learning_rate": 0.00029654654654654654,
|
| 2126 |
+
"loss": 0.6378,
|
| 2127 |
+
"step": 3450
|
| 2128 |
+
},
|
| 2129 |
+
{
|
| 2130 |
+
"epoch": 4.67,
|
| 2131 |
+
"learning_rate": 0.0002957957957957958,
|
| 2132 |
+
"loss": 0.6456,
|
| 2133 |
+
"step": 3460
|
| 2134 |
+
},
|
| 2135 |
+
{
|
| 2136 |
+
"epoch": 4.68,
|
| 2137 |
+
"learning_rate": 0.000295045045045045,
|
| 2138 |
+
"loss": 0.6649,
|
| 2139 |
+
"step": 3470
|
| 2140 |
+
},
|
| 2141 |
+
{
|
| 2142 |
+
"epoch": 4.7,
|
| 2143 |
+
"learning_rate": 0.0002942942942942943,
|
| 2144 |
+
"loss": 0.6672,
|
| 2145 |
+
"step": 3480
|
| 2146 |
+
},
|
| 2147 |
+
{
|
| 2148 |
+
"epoch": 4.71,
|
| 2149 |
+
"learning_rate": 0.00029354354354354357,
|
| 2150 |
+
"loss": 0.6567,
|
| 2151 |
+
"step": 3490
|
| 2152 |
+
},
|
| 2153 |
+
{
|
| 2154 |
+
"epoch": 4.72,
|
| 2155 |
+
"learning_rate": 0.0002927927927927928,
|
| 2156 |
+
"loss": 0.6411,
|
| 2157 |
+
"step": 3500
|
| 2158 |
+
},
|
| 2159 |
+
{
|
| 2160 |
+
"epoch": 4.74,
|
| 2161 |
+
"learning_rate": 0.00029204204204204206,
|
| 2162 |
+
"loss": 0.6502,
|
| 2163 |
+
"step": 3510
|
| 2164 |
+
},
|
| 2165 |
+
{
|
| 2166 |
+
"epoch": 4.75,
|
| 2167 |
+
"learning_rate": 0.00029129129129129127,
|
| 2168 |
+
"loss": 0.6637,
|
| 2169 |
+
"step": 3520
|
| 2170 |
+
},
|
| 2171 |
+
{
|
| 2172 |
+
"epoch": 4.77,
|
| 2173 |
+
"learning_rate": 0.00029054054054054054,
|
| 2174 |
+
"loss": 0.6339,
|
| 2175 |
+
"step": 3530
|
| 2176 |
+
},
|
| 2177 |
+
{
|
| 2178 |
+
"epoch": 4.78,
|
| 2179 |
+
"learning_rate": 0.0002897897897897898,
|
| 2180 |
+
"loss": 0.5812,
|
| 2181 |
+
"step": 3540
|
| 2182 |
+
},
|
| 2183 |
+
{
|
| 2184 |
+
"epoch": 4.79,
|
| 2185 |
+
"learning_rate": 0.00028903903903903903,
|
| 2186 |
+
"loss": 0.6396,
|
| 2187 |
+
"step": 3550
|
| 2188 |
+
},
|
| 2189 |
+
{
|
| 2190 |
+
"epoch": 4.81,
|
| 2191 |
+
"learning_rate": 0.0002882882882882883,
|
| 2192 |
+
"loss": 0.6208,
|
| 2193 |
+
"step": 3560
|
| 2194 |
+
},
|
| 2195 |
+
{
|
| 2196 |
+
"epoch": 4.82,
|
| 2197 |
+
"learning_rate": 0.0002875375375375375,
|
| 2198 |
+
"loss": 0.6381,
|
| 2199 |
+
"step": 3570
|
| 2200 |
+
},
|
| 2201 |
+
{
|
| 2202 |
+
"epoch": 4.83,
|
| 2203 |
+
"learning_rate": 0.0002867867867867868,
|
| 2204 |
+
"loss": 0.691,
|
| 2205 |
+
"step": 3580
|
| 2206 |
+
},
|
| 2207 |
+
{
|
| 2208 |
+
"epoch": 4.85,
|
| 2209 |
+
"learning_rate": 0.00028603603603603606,
|
| 2210 |
+
"loss": 0.7211,
|
| 2211 |
+
"step": 3590
|
| 2212 |
+
},
|
| 2213 |
+
{
|
| 2214 |
+
"epoch": 4.86,
|
| 2215 |
+
"learning_rate": 0.0002852852852852853,
|
| 2216 |
+
"loss": 0.616,
|
| 2217 |
+
"step": 3600
|
| 2218 |
+
},
|
| 2219 |
+
{
|
| 2220 |
+
"epoch": 4.87,
|
| 2221 |
+
"learning_rate": 0.00028453453453453455,
|
| 2222 |
+
"loss": 0.5619,
|
| 2223 |
+
"step": 3610
|
| 2224 |
+
},
|
| 2225 |
+
{
|
| 2226 |
+
"epoch": 4.89,
|
| 2227 |
+
"learning_rate": 0.00028378378378378377,
|
| 2228 |
+
"loss": 0.5765,
|
| 2229 |
+
"step": 3620
|
| 2230 |
+
},
|
| 2231 |
+
{
|
| 2232 |
+
"epoch": 4.9,
|
| 2233 |
+
"learning_rate": 0.00028303303303303304,
|
| 2234 |
+
"loss": 0.6436,
|
| 2235 |
+
"step": 3630
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 4.91,
|
| 2239 |
+
"learning_rate": 0.0002822822822822823,
|
| 2240 |
+
"loss": 0.6864,
|
| 2241 |
+
"step": 3640
|
| 2242 |
+
},
|
| 2243 |
+
{
|
| 2244 |
+
"epoch": 4.93,
|
| 2245 |
+
"learning_rate": 0.00028153153153153153,
|
| 2246 |
+
"loss": 0.6217,
|
| 2247 |
+
"step": 3650
|
| 2248 |
+
},
|
| 2249 |
+
{
|
| 2250 |
+
"epoch": 4.94,
|
| 2251 |
+
"learning_rate": 0.0002807807807807808,
|
| 2252 |
+
"loss": 0.5852,
|
| 2253 |
+
"step": 3660
|
| 2254 |
+
},
|
| 2255 |
+
{
|
| 2256 |
+
"epoch": 4.95,
|
| 2257 |
+
"learning_rate": 0.00028003003003003,
|
| 2258 |
+
"loss": 0.6185,
|
| 2259 |
+
"step": 3670
|
| 2260 |
+
},
|
| 2261 |
+
{
|
| 2262 |
+
"epoch": 4.97,
|
| 2263 |
+
"learning_rate": 0.0002792792792792793,
|
| 2264 |
+
"loss": 0.6803,
|
| 2265 |
+
"step": 3680
|
| 2266 |
+
},
|
| 2267 |
+
{
|
| 2268 |
+
"epoch": 4.98,
|
| 2269 |
+
"learning_rate": 0.00027852852852852856,
|
| 2270 |
+
"loss": 0.5834,
|
| 2271 |
+
"step": 3690
|
| 2272 |
+
},
|
| 2273 |
+
{
|
| 2274 |
+
"epoch": 4.99,
|
| 2275 |
+
"learning_rate": 0.0002777777777777778,
|
| 2276 |
+
"loss": 0.6357,
|
| 2277 |
+
"step": 3700
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 5.0,
|
| 2281 |
+
"eval_accuracy": 0.8704740948189638,
|
| 2282 |
+
"eval_f1": 0.8689097751341385,
|
| 2283 |
+
"eval_loss": 0.44387924671173096,
|
| 2284 |
+
"eval_precision": 0.8761662335945799,
|
| 2285 |
+
"eval_recall": 0.8693479074500065,
|
| 2286 |
+
"eval_runtime": 191.3953,
|
| 2287 |
+
"eval_samples_per_second": 52.237,
|
| 2288 |
+
"eval_steps_per_second": 1.938,
|
| 2289 |
+
"step": 3703
|
| 2290 |
+
},
|
| 2291 |
+
{
|
| 2292 |
+
"epoch": 5.01,
|
| 2293 |
+
"learning_rate": 0.00027702702702702705,
|
| 2294 |
+
"loss": 0.5605,
|
| 2295 |
+
"step": 3710
|
| 2296 |
+
},
|
| 2297 |
+
{
|
| 2298 |
+
"epoch": 5.02,
|
| 2299 |
+
"learning_rate": 0.00027627627627627627,
|
| 2300 |
+
"loss": 0.5468,
|
| 2301 |
+
"step": 3720
|
| 2302 |
+
},
|
| 2303 |
+
{
|
| 2304 |
+
"epoch": 5.04,
|
| 2305 |
+
"learning_rate": 0.00027552552552552554,
|
| 2306 |
+
"loss": 0.5749,
|
| 2307 |
+
"step": 3730
|
| 2308 |
+
},
|
| 2309 |
+
{
|
| 2310 |
+
"epoch": 5.05,
|
| 2311 |
+
"learning_rate": 0.0002747747747747748,
|
| 2312 |
+
"loss": 0.53,
|
| 2313 |
+
"step": 3740
|
| 2314 |
+
},
|
| 2315 |
+
{
|
| 2316 |
+
"epoch": 5.06,
|
| 2317 |
+
"learning_rate": 0.000274024024024024,
|
| 2318 |
+
"loss": 0.5308,
|
| 2319 |
+
"step": 3750
|
| 2320 |
+
},
|
| 2321 |
+
{
|
| 2322 |
+
"epoch": 5.08,
|
| 2323 |
+
"learning_rate": 0.0002732732732732733,
|
| 2324 |
+
"loss": 0.5127,
|
| 2325 |
+
"step": 3760
|
| 2326 |
+
},
|
| 2327 |
+
{
|
| 2328 |
+
"epoch": 5.09,
|
| 2329 |
+
"learning_rate": 0.0002725225225225225,
|
| 2330 |
+
"loss": 0.5424,
|
| 2331 |
+
"step": 3770
|
| 2332 |
+
},
|
| 2333 |
+
{
|
| 2334 |
+
"epoch": 5.1,
|
| 2335 |
+
"learning_rate": 0.00027177177177177173,
|
| 2336 |
+
"loss": 0.5608,
|
| 2337 |
+
"step": 3780
|
| 2338 |
+
},
|
| 2339 |
+
{
|
| 2340 |
+
"epoch": 5.12,
|
| 2341 |
+
"learning_rate": 0.00027102102102102106,
|
| 2342 |
+
"loss": 0.621,
|
| 2343 |
+
"step": 3790
|
| 2344 |
+
},
|
| 2345 |
+
{
|
| 2346 |
+
"epoch": 5.13,
|
| 2347 |
+
"learning_rate": 0.0002702702702702703,
|
| 2348 |
+
"loss": 0.5018,
|
| 2349 |
+
"step": 3800
|
| 2350 |
+
},
|
| 2351 |
+
{
|
| 2352 |
+
"epoch": 5.14,
|
| 2353 |
+
"learning_rate": 0.00026951951951951955,
|
| 2354 |
+
"loss": 0.4879,
|
| 2355 |
+
"step": 3810
|
| 2356 |
+
},
|
| 2357 |
+
{
|
| 2358 |
+
"epoch": 5.16,
|
| 2359 |
+
"learning_rate": 0.00026876876876876876,
|
| 2360 |
+
"loss": 0.4991,
|
| 2361 |
+
"step": 3820
|
| 2362 |
+
},
|
| 2363 |
+
{
|
| 2364 |
+
"epoch": 5.17,
|
| 2365 |
+
"learning_rate": 0.000268018018018018,
|
| 2366 |
+
"loss": 0.5633,
|
| 2367 |
+
"step": 3830
|
| 2368 |
+
},
|
| 2369 |
+
{
|
| 2370 |
+
"epoch": 5.18,
|
| 2371 |
+
"learning_rate": 0.0002672672672672673,
|
| 2372 |
+
"loss": 0.6233,
|
| 2373 |
+
"step": 3840
|
| 2374 |
+
},
|
| 2375 |
+
{
|
| 2376 |
+
"epoch": 5.2,
|
| 2377 |
+
"learning_rate": 0.0002665165165165165,
|
| 2378 |
+
"loss": 0.511,
|
| 2379 |
+
"step": 3850
|
| 2380 |
+
},
|
| 2381 |
+
{
|
| 2382 |
+
"epoch": 5.21,
|
| 2383 |
+
"learning_rate": 0.0002657657657657658,
|
| 2384 |
+
"loss": 0.5036,
|
| 2385 |
+
"step": 3860
|
| 2386 |
+
},
|
| 2387 |
+
{
|
| 2388 |
+
"epoch": 5.22,
|
| 2389 |
+
"learning_rate": 0.000265015015015015,
|
| 2390 |
+
"loss": 0.5826,
|
| 2391 |
+
"step": 3870
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 5.24,
|
| 2395 |
+
"learning_rate": 0.00026426426426426423,
|
| 2396 |
+
"loss": 0.5605,
|
| 2397 |
+
"step": 3880
|
| 2398 |
+
},
|
| 2399 |
+
{
|
| 2400 |
+
"epoch": 5.25,
|
| 2401 |
+
"learning_rate": 0.0002635135135135135,
|
| 2402 |
+
"loss": 0.6443,
|
| 2403 |
+
"step": 3890
|
| 2404 |
+
},
|
| 2405 |
+
{
|
| 2406 |
+
"epoch": 5.26,
|
| 2407 |
+
"learning_rate": 0.00026276276276276277,
|
| 2408 |
+
"loss": 0.582,
|
| 2409 |
+
"step": 3900
|
| 2410 |
+
},
|
| 2411 |
+
{
|
| 2412 |
+
"epoch": 5.28,
|
| 2413 |
+
"learning_rate": 0.00026201201201201204,
|
| 2414 |
+
"loss": 0.6232,
|
| 2415 |
+
"step": 3910
|
| 2416 |
+
},
|
| 2417 |
+
{
|
| 2418 |
+
"epoch": 5.29,
|
| 2419 |
+
"learning_rate": 0.00026126126126126126,
|
| 2420 |
+
"loss": 0.5763,
|
| 2421 |
+
"step": 3920
|
| 2422 |
+
},
|
| 2423 |
+
{
|
| 2424 |
+
"epoch": 5.31,
|
| 2425 |
+
"learning_rate": 0.0002605105105105105,
|
| 2426 |
+
"loss": 0.6183,
|
| 2427 |
+
"step": 3930
|
| 2428 |
+
},
|
| 2429 |
+
{
|
| 2430 |
+
"epoch": 5.32,
|
| 2431 |
+
"learning_rate": 0.00025975975975975975,
|
| 2432 |
+
"loss": 0.5533,
|
| 2433 |
+
"step": 3940
|
| 2434 |
+
},
|
| 2435 |
+
{
|
| 2436 |
+
"epoch": 5.33,
|
| 2437 |
+
"learning_rate": 0.000259009009009009,
|
| 2438 |
+
"loss": 0.5633,
|
| 2439 |
+
"step": 3950
|
| 2440 |
+
},
|
| 2441 |
+
{
|
| 2442 |
+
"epoch": 5.35,
|
| 2443 |
+
"learning_rate": 0.0002582582582582583,
|
| 2444 |
+
"loss": 0.544,
|
| 2445 |
+
"step": 3960
|
| 2446 |
+
},
|
| 2447 |
+
{
|
| 2448 |
+
"epoch": 5.36,
|
| 2449 |
+
"learning_rate": 0.0002575075075075075,
|
| 2450 |
+
"loss": 0.5882,
|
| 2451 |
+
"step": 3970
|
| 2452 |
+
},
|
| 2453 |
+
{
|
| 2454 |
+
"epoch": 5.37,
|
| 2455 |
+
"learning_rate": 0.0002567567567567567,
|
| 2456 |
+
"loss": 0.5974,
|
| 2457 |
+
"step": 3980
|
| 2458 |
+
},
|
| 2459 |
+
{
|
| 2460 |
+
"epoch": 5.39,
|
| 2461 |
+
"learning_rate": 0.000256006006006006,
|
| 2462 |
+
"loss": 0.557,
|
| 2463 |
+
"step": 3990
|
| 2464 |
+
},
|
| 2465 |
+
{
|
| 2466 |
+
"epoch": 5.4,
|
| 2467 |
+
"learning_rate": 0.00025525525525525527,
|
| 2468 |
+
"loss": 0.5013,
|
| 2469 |
+
"step": 4000
|
| 2470 |
+
},
|
| 2471 |
+
{
|
| 2472 |
+
"epoch": 5.41,
|
| 2473 |
+
"learning_rate": 0.00025450450450450454,
|
| 2474 |
+
"loss": 0.5785,
|
| 2475 |
+
"step": 4010
|
| 2476 |
+
},
|
| 2477 |
+
{
|
| 2478 |
+
"epoch": 5.43,
|
| 2479 |
+
"learning_rate": 0.00025375375375375376,
|
| 2480 |
+
"loss": 0.558,
|
| 2481 |
+
"step": 4020
|
| 2482 |
+
},
|
| 2483 |
+
{
|
| 2484 |
+
"epoch": 5.44,
|
| 2485 |
+
"learning_rate": 0.000253003003003003,
|
| 2486 |
+
"loss": 0.5429,
|
| 2487 |
+
"step": 4030
|
| 2488 |
+
},
|
| 2489 |
+
{
|
| 2490 |
+
"epoch": 5.45,
|
| 2491 |
+
"learning_rate": 0.00025225225225225225,
|
| 2492 |
+
"loss": 0.5667,
|
| 2493 |
+
"step": 4040
|
| 2494 |
+
},
|
| 2495 |
+
{
|
| 2496 |
+
"epoch": 5.47,
|
| 2497 |
+
"learning_rate": 0.0002515015015015015,
|
| 2498 |
+
"loss": 0.5386,
|
| 2499 |
+
"step": 4050
|
| 2500 |
+
},
|
| 2501 |
+
{
|
| 2502 |
+
"epoch": 5.48,
|
| 2503 |
+
"learning_rate": 0.0002507507507507508,
|
| 2504 |
+
"loss": 0.6074,
|
| 2505 |
+
"step": 4060
|
| 2506 |
+
},
|
| 2507 |
+
{
|
| 2508 |
+
"epoch": 5.49,
|
| 2509 |
+
"learning_rate": 0.00025,
|
| 2510 |
+
"loss": 0.5583,
|
| 2511 |
+
"step": 4070
|
| 2512 |
+
},
|
| 2513 |
+
{
|
| 2514 |
+
"epoch": 5.51,
|
| 2515 |
+
"learning_rate": 0.0002492492492492492,
|
| 2516 |
+
"loss": 0.4637,
|
| 2517 |
+
"step": 4080
|
| 2518 |
+
},
|
| 2519 |
+
{
|
| 2520 |
+
"epoch": 5.52,
|
| 2521 |
+
"learning_rate": 0.0002484984984984985,
|
| 2522 |
+
"loss": 0.5877,
|
| 2523 |
+
"step": 4090
|
| 2524 |
+
},
|
| 2525 |
+
{
|
| 2526 |
+
"epoch": 5.53,
|
| 2527 |
+
"learning_rate": 0.00024774774774774777,
|
| 2528 |
+
"loss": 0.4403,
|
| 2529 |
+
"step": 4100
|
| 2530 |
+
},
|
| 2531 |
+
{
|
| 2532 |
+
"epoch": 5.55,
|
| 2533 |
+
"learning_rate": 0.00024699699699699704,
|
| 2534 |
+
"loss": 0.5379,
|
| 2535 |
+
"step": 4110
|
| 2536 |
+
},
|
| 2537 |
+
{
|
| 2538 |
+
"epoch": 5.56,
|
| 2539 |
+
"learning_rate": 0.00024624624624624625,
|
| 2540 |
+
"loss": 0.5562,
|
| 2541 |
+
"step": 4120
|
| 2542 |
+
},
|
| 2543 |
+
{
|
| 2544 |
+
"epoch": 5.58,
|
| 2545 |
+
"learning_rate": 0.00024549549549549547,
|
| 2546 |
+
"loss": 0.5481,
|
| 2547 |
+
"step": 4130
|
| 2548 |
+
},
|
| 2549 |
+
{
|
| 2550 |
+
"epoch": 5.59,
|
| 2551 |
+
"learning_rate": 0.00024474474474474474,
|
| 2552 |
+
"loss": 0.5422,
|
| 2553 |
+
"step": 4140
|
| 2554 |
+
},
|
| 2555 |
+
{
|
| 2556 |
+
"epoch": 5.6,
|
| 2557 |
+
"learning_rate": 0.000243993993993994,
|
| 2558 |
+
"loss": 0.5545,
|
| 2559 |
+
"step": 4150
|
| 2560 |
+
},
|
| 2561 |
+
{
|
| 2562 |
+
"epoch": 5.62,
|
| 2563 |
+
"learning_rate": 0.00024324324324324326,
|
| 2564 |
+
"loss": 0.5128,
|
| 2565 |
+
"step": 4160
|
| 2566 |
+
},
|
| 2567 |
+
{
|
| 2568 |
+
"epoch": 5.63,
|
| 2569 |
+
"learning_rate": 0.0002424924924924925,
|
| 2570 |
+
"loss": 0.5653,
|
| 2571 |
+
"step": 4170
|
| 2572 |
+
},
|
| 2573 |
+
{
|
| 2574 |
+
"epoch": 5.64,
|
| 2575 |
+
"learning_rate": 0.00024174174174174175,
|
| 2576 |
+
"loss": 0.5782,
|
| 2577 |
+
"step": 4180
|
| 2578 |
+
},
|
| 2579 |
+
{
|
| 2580 |
+
"epoch": 5.66,
|
| 2581 |
+
"learning_rate": 0.00024099099099099102,
|
| 2582 |
+
"loss": 0.5709,
|
| 2583 |
+
"step": 4190
|
| 2584 |
+
},
|
| 2585 |
+
{
|
| 2586 |
+
"epoch": 5.67,
|
| 2587 |
+
"learning_rate": 0.00024024024024024023,
|
| 2588 |
+
"loss": 0.5307,
|
| 2589 |
+
"step": 4200
|
| 2590 |
+
},
|
| 2591 |
+
{
|
| 2592 |
+
"epoch": 5.68,
|
| 2593 |
+
"learning_rate": 0.00023948948948948948,
|
| 2594 |
+
"loss": 0.569,
|
| 2595 |
+
"step": 4210
|
| 2596 |
+
},
|
| 2597 |
+
{
|
| 2598 |
+
"epoch": 5.7,
|
| 2599 |
+
"learning_rate": 0.00023873873873873875,
|
| 2600 |
+
"loss": 0.5939,
|
| 2601 |
+
"step": 4220
|
| 2602 |
+
},
|
| 2603 |
+
{
|
| 2604 |
+
"epoch": 5.71,
|
| 2605 |
+
"learning_rate": 0.000237987987987988,
|
| 2606 |
+
"loss": 0.5127,
|
| 2607 |
+
"step": 4230
|
| 2608 |
+
},
|
| 2609 |
+
{
|
| 2610 |
+
"epoch": 5.72,
|
| 2611 |
+
"learning_rate": 0.00023723723723723724,
|
| 2612 |
+
"loss": 0.5174,
|
| 2613 |
+
"step": 4240
|
| 2614 |
+
},
|
| 2615 |
+
{
|
| 2616 |
+
"epoch": 5.74,
|
| 2617 |
+
"learning_rate": 0.00023648648648648648,
|
| 2618 |
+
"loss": 0.558,
|
| 2619 |
+
"step": 4250
|
| 2620 |
+
},
|
| 2621 |
+
{
|
| 2622 |
+
"epoch": 5.75,
|
| 2623 |
+
"learning_rate": 0.00023573573573573573,
|
| 2624 |
+
"loss": 0.5006,
|
| 2625 |
+
"step": 4260
|
| 2626 |
+
},
|
| 2627 |
+
{
|
| 2628 |
+
"epoch": 5.76,
|
| 2629 |
+
"learning_rate": 0.000234984984984985,
|
| 2630 |
+
"loss": 0.6208,
|
| 2631 |
+
"step": 4270
|
| 2632 |
+
},
|
| 2633 |
+
{
|
| 2634 |
+
"epoch": 5.78,
|
| 2635 |
+
"learning_rate": 0.00023423423423423424,
|
| 2636 |
+
"loss": 0.554,
|
| 2637 |
+
"step": 4280
|
| 2638 |
+
},
|
| 2639 |
+
{
|
| 2640 |
+
"epoch": 5.79,
|
| 2641 |
+
"learning_rate": 0.0002334834834834835,
|
| 2642 |
+
"loss": 0.5183,
|
| 2643 |
+
"step": 4290
|
| 2644 |
+
},
|
| 2645 |
+
{
|
| 2646 |
+
"epoch": 5.8,
|
| 2647 |
+
"learning_rate": 0.00023273273273273273,
|
| 2648 |
+
"loss": 0.5545,
|
| 2649 |
+
"step": 4300
|
| 2650 |
+
},
|
| 2651 |
+
{
|
| 2652 |
+
"epoch": 5.82,
|
| 2653 |
+
"learning_rate": 0.00023198198198198198,
|
| 2654 |
+
"loss": 0.5011,
|
| 2655 |
+
"step": 4310
|
| 2656 |
+
},
|
| 2657 |
+
{
|
| 2658 |
+
"epoch": 5.83,
|
| 2659 |
+
"learning_rate": 0.00023123123123123125,
|
| 2660 |
+
"loss": 0.5286,
|
| 2661 |
+
"step": 4320
|
| 2662 |
+
},
|
| 2663 |
+
{
|
| 2664 |
+
"epoch": 5.85,
|
| 2665 |
+
"learning_rate": 0.0002304804804804805,
|
| 2666 |
+
"loss": 0.4974,
|
| 2667 |
+
"step": 4330
|
| 2668 |
+
},
|
| 2669 |
+
{
|
| 2670 |
+
"epoch": 5.86,
|
| 2671 |
+
"learning_rate": 0.00022972972972972974,
|
| 2672 |
+
"loss": 0.576,
|
| 2673 |
+
"step": 4340
|
| 2674 |
+
},
|
| 2675 |
+
{
|
| 2676 |
+
"epoch": 5.87,
|
| 2677 |
+
"learning_rate": 0.00022897897897897898,
|
| 2678 |
+
"loss": 0.4924,
|
| 2679 |
+
"step": 4350
|
| 2680 |
+
},
|
| 2681 |
+
{
|
| 2682 |
+
"epoch": 5.89,
|
| 2683 |
+
"learning_rate": 0.00022822822822822822,
|
| 2684 |
+
"loss": 0.5501,
|
| 2685 |
+
"step": 4360
|
| 2686 |
+
},
|
| 2687 |
+
{
|
| 2688 |
+
"epoch": 5.9,
|
| 2689 |
+
"learning_rate": 0.0002274774774774775,
|
| 2690 |
+
"loss": 0.5307,
|
| 2691 |
+
"step": 4370
|
| 2692 |
+
},
|
| 2693 |
+
{
|
| 2694 |
+
"epoch": 5.91,
|
| 2695 |
+
"learning_rate": 0.00022672672672672674,
|
| 2696 |
+
"loss": 0.5161,
|
| 2697 |
+
"step": 4380
|
| 2698 |
+
},
|
| 2699 |
+
{
|
| 2700 |
+
"epoch": 5.93,
|
| 2701 |
+
"learning_rate": 0.00022597597597597596,
|
| 2702 |
+
"loss": 0.5153,
|
| 2703 |
+
"step": 4390
|
| 2704 |
+
},
|
| 2705 |
+
{
|
| 2706 |
+
"epoch": 5.94,
|
| 2707 |
+
"learning_rate": 0.00022522522522522523,
|
| 2708 |
+
"loss": 0.5222,
|
| 2709 |
+
"step": 4400
|
| 2710 |
+
},
|
| 2711 |
+
{
|
| 2712 |
+
"epoch": 5.95,
|
| 2713 |
+
"learning_rate": 0.00022447447447447447,
|
| 2714 |
+
"loss": 0.5144,
|
| 2715 |
+
"step": 4410
|
| 2716 |
+
},
|
| 2717 |
+
{
|
| 2718 |
+
"epoch": 5.97,
|
| 2719 |
+
"learning_rate": 0.00022372372372372374,
|
| 2720 |
+
"loss": 0.5002,
|
| 2721 |
+
"step": 4420
|
| 2722 |
+
},
|
| 2723 |
+
{
|
| 2724 |
+
"epoch": 5.98,
|
| 2725 |
+
"learning_rate": 0.000222972972972973,
|
| 2726 |
+
"loss": 0.4556,
|
| 2727 |
+
"step": 4430
|
| 2728 |
+
},
|
| 2729 |
+
{
|
| 2730 |
+
"epoch": 5.99,
|
| 2731 |
+
"learning_rate": 0.0002222222222222222,
|
| 2732 |
+
"loss": 0.5657,
|
| 2733 |
+
"step": 4440
|
| 2734 |
+
},
|
| 2735 |
+
{
|
| 2736 |
+
"epoch": 6.0,
|
| 2737 |
+
"eval_accuracy": 0.8890778155631126,
|
| 2738 |
+
"eval_f1": 0.8884475854159729,
|
| 2739 |
+
"eval_loss": 0.405392587184906,
|
| 2740 |
+
"eval_precision": 0.8929255500739524,
|
| 2741 |
+
"eval_recall": 0.8885232893443815,
|
| 2742 |
+
"eval_runtime": 191.4068,
|
| 2743 |
+
"eval_samples_per_second": 52.234,
|
| 2744 |
+
"eval_steps_per_second": 1.938,
|
| 2745 |
+
"step": 4444
|
| 2746 |
+
},
|
| 2747 |
+
{
|
| 2748 |
+
"epoch": 6.01,
|
| 2749 |
+
"learning_rate": 0.00022147147147147148,
|
| 2750 |
+
"loss": 0.5383,
|
| 2751 |
+
"step": 4450
|
| 2752 |
+
},
|
| 2753 |
+
{
|
| 2754 |
+
"epoch": 6.02,
|
| 2755 |
+
"learning_rate": 0.00022072072072072072,
|
| 2756 |
+
"loss": 0.444,
|
| 2757 |
+
"step": 4460
|
| 2758 |
+
},
|
| 2759 |
+
{
|
| 2760 |
+
"epoch": 6.03,
|
| 2761 |
+
"learning_rate": 0.00021996996996997,
|
| 2762 |
+
"loss": 0.4961,
|
| 2763 |
+
"step": 4470
|
| 2764 |
+
},
|
| 2765 |
+
{
|
| 2766 |
+
"epoch": 6.05,
|
| 2767 |
+
"learning_rate": 0.00021921921921921924,
|
| 2768 |
+
"loss": 0.5372,
|
| 2769 |
+
"step": 4480
|
| 2770 |
+
},
|
| 2771 |
+
{
|
| 2772 |
+
"epoch": 6.06,
|
| 2773 |
+
"learning_rate": 0.00021846846846846845,
|
| 2774 |
+
"loss": 0.4424,
|
| 2775 |
+
"step": 4490
|
| 2776 |
+
},
|
| 2777 |
+
{
|
| 2778 |
+
"epoch": 6.07,
|
| 2779 |
+
"learning_rate": 0.00021771771771771773,
|
| 2780 |
+
"loss": 0.3961,
|
| 2781 |
+
"step": 4500
|
| 2782 |
+
},
|
| 2783 |
+
{
|
| 2784 |
+
"epoch": 6.09,
|
| 2785 |
+
"learning_rate": 0.00021696696696696697,
|
| 2786 |
+
"loss": 0.4882,
|
| 2787 |
+
"step": 4510
|
| 2788 |
+
},
|
| 2789 |
+
{
|
| 2790 |
+
"epoch": 6.1,
|
| 2791 |
+
"learning_rate": 0.00021621621621621624,
|
| 2792 |
+
"loss": 0.4517,
|
| 2793 |
+
"step": 4520
|
| 2794 |
+
},
|
| 2795 |
+
{
|
| 2796 |
+
"epoch": 6.12,
|
| 2797 |
+
"learning_rate": 0.00021546546546546546,
|
| 2798 |
+
"loss": 0.4742,
|
| 2799 |
+
"step": 4530
|
| 2800 |
+
},
|
| 2801 |
+
{
|
| 2802 |
+
"epoch": 6.13,
|
| 2803 |
+
"learning_rate": 0.0002147147147147147,
|
| 2804 |
+
"loss": 0.5285,
|
| 2805 |
+
"step": 4540
|
| 2806 |
+
},
|
| 2807 |
+
{
|
| 2808 |
+
"epoch": 6.14,
|
| 2809 |
+
"learning_rate": 0.00021396396396396397,
|
| 2810 |
+
"loss": 0.4771,
|
| 2811 |
+
"step": 4550
|
| 2812 |
+
},
|
| 2813 |
+
{
|
| 2814 |
+
"epoch": 6.16,
|
| 2815 |
+
"learning_rate": 0.00021321321321321322,
|
| 2816 |
+
"loss": 0.4278,
|
| 2817 |
+
"step": 4560
|
| 2818 |
+
},
|
| 2819 |
+
{
|
| 2820 |
+
"epoch": 6.17,
|
| 2821 |
+
"learning_rate": 0.0002124624624624625,
|
| 2822 |
+
"loss": 0.5205,
|
| 2823 |
+
"step": 4570
|
| 2824 |
+
},
|
| 2825 |
+
{
|
| 2826 |
+
"epoch": 6.18,
|
| 2827 |
+
"learning_rate": 0.0002117117117117117,
|
| 2828 |
+
"loss": 0.5019,
|
| 2829 |
+
"step": 4580
|
| 2830 |
+
},
|
| 2831 |
+
{
|
| 2832 |
+
"epoch": 6.2,
|
| 2833 |
+
"learning_rate": 0.00021096096096096095,
|
| 2834 |
+
"loss": 0.4504,
|
| 2835 |
+
"step": 4590
|
| 2836 |
+
},
|
| 2837 |
+
{
|
| 2838 |
+
"epoch": 6.21,
|
| 2839 |
+
"learning_rate": 0.00021021021021021022,
|
| 2840 |
+
"loss": 0.411,
|
| 2841 |
+
"step": 4600
|
| 2842 |
+
},
|
| 2843 |
+
{
|
| 2844 |
+
"epoch": 6.22,
|
| 2845 |
+
"learning_rate": 0.00020945945945945947,
|
| 2846 |
+
"loss": 0.5114,
|
| 2847 |
+
"step": 4610
|
| 2848 |
+
},
|
| 2849 |
+
{
|
| 2850 |
+
"epoch": 6.24,
|
| 2851 |
+
"learning_rate": 0.0002087087087087087,
|
| 2852 |
+
"loss": 0.4836,
|
| 2853 |
+
"step": 4620
|
| 2854 |
+
},
|
| 2855 |
+
{
|
| 2856 |
+
"epoch": 6.25,
|
| 2857 |
+
"learning_rate": 0.00020795795795795796,
|
| 2858 |
+
"loss": 0.4559,
|
| 2859 |
+
"step": 4630
|
| 2860 |
+
},
|
| 2861 |
+
{
|
| 2862 |
+
"epoch": 6.26,
|
| 2863 |
+
"learning_rate": 0.0002072072072072072,
|
| 2864 |
+
"loss": 0.4264,
|
| 2865 |
+
"step": 4640
|
| 2866 |
+
},
|
| 2867 |
+
{
|
| 2868 |
+
"epoch": 6.28,
|
| 2869 |
+
"learning_rate": 0.00020645645645645647,
|
| 2870 |
+
"loss": 0.4078,
|
| 2871 |
+
"step": 4650
|
| 2872 |
+
},
|
| 2873 |
+
{
|
| 2874 |
+
"epoch": 6.29,
|
| 2875 |
+
"learning_rate": 0.00020570570570570572,
|
| 2876 |
+
"loss": 0.451,
|
| 2877 |
+
"step": 4660
|
| 2878 |
+
},
|
| 2879 |
+
{
|
| 2880 |
+
"epoch": 6.3,
|
| 2881 |
+
"learning_rate": 0.00020495495495495496,
|
| 2882 |
+
"loss": 0.407,
|
| 2883 |
+
"step": 4670
|
| 2884 |
+
},
|
| 2885 |
+
{
|
| 2886 |
+
"epoch": 6.32,
|
| 2887 |
+
"learning_rate": 0.0002042042042042042,
|
| 2888 |
+
"loss": 0.412,
|
| 2889 |
+
"step": 4680
|
| 2890 |
+
},
|
| 2891 |
+
{
|
| 2892 |
+
"epoch": 6.33,
|
| 2893 |
+
"learning_rate": 0.00020345345345345345,
|
| 2894 |
+
"loss": 0.502,
|
| 2895 |
+
"step": 4690
|
| 2896 |
+
},
|
| 2897 |
+
{
|
| 2898 |
+
"epoch": 6.34,
|
| 2899 |
+
"learning_rate": 0.00020270270270270272,
|
| 2900 |
+
"loss": 0.4881,
|
| 2901 |
+
"step": 4700
|
| 2902 |
+
},
|
| 2903 |
+
{
|
| 2904 |
+
"epoch": 6.36,
|
| 2905 |
+
"learning_rate": 0.00020195195195195196,
|
| 2906 |
+
"loss": 0.4685,
|
| 2907 |
+
"step": 4710
|
| 2908 |
+
},
|
| 2909 |
+
{
|
| 2910 |
+
"epoch": 6.37,
|
| 2911 |
+
"learning_rate": 0.0002012012012012012,
|
| 2912 |
+
"loss": 0.4531,
|
| 2913 |
+
"step": 4720
|
| 2914 |
+
},
|
| 2915 |
+
{
|
| 2916 |
+
"epoch": 6.39,
|
| 2917 |
+
"learning_rate": 0.00020045045045045045,
|
| 2918 |
+
"loss": 0.3962,
|
| 2919 |
+
"step": 4730
|
| 2920 |
+
},
|
| 2921 |
+
{
|
| 2922 |
+
"epoch": 6.4,
|
| 2923 |
+
"learning_rate": 0.0001996996996996997,
|
| 2924 |
+
"loss": 0.3741,
|
| 2925 |
+
"step": 4740
|
| 2926 |
+
},
|
| 2927 |
+
{
|
| 2928 |
+
"epoch": 6.41,
|
| 2929 |
+
"learning_rate": 0.00019894894894894897,
|
| 2930 |
+
"loss": 0.4502,
|
| 2931 |
+
"step": 4750
|
| 2932 |
+
},
|
| 2933 |
+
{
|
| 2934 |
+
"epoch": 6.43,
|
| 2935 |
+
"learning_rate": 0.0001981981981981982,
|
| 2936 |
+
"loss": 0.4775,
|
| 2937 |
+
"step": 4760
|
| 2938 |
+
},
|
| 2939 |
+
{
|
| 2940 |
+
"epoch": 6.44,
|
| 2941 |
+
"learning_rate": 0.00019744744744744743,
|
| 2942 |
+
"loss": 0.5319,
|
| 2943 |
+
"step": 4770
|
| 2944 |
+
},
|
| 2945 |
+
{
|
| 2946 |
+
"epoch": 6.45,
|
| 2947 |
+
"learning_rate": 0.0001966966966966967,
|
| 2948 |
+
"loss": 0.5363,
|
| 2949 |
+
"step": 4780
|
| 2950 |
+
},
|
| 2951 |
+
{
|
| 2952 |
+
"epoch": 6.47,
|
| 2953 |
+
"learning_rate": 0.00019594594594594594,
|
| 2954 |
+
"loss": 0.4511,
|
| 2955 |
+
"step": 4790
|
| 2956 |
+
},
|
| 2957 |
+
{
|
| 2958 |
+
"epoch": 6.48,
|
| 2959 |
+
"learning_rate": 0.0001951951951951952,
|
| 2960 |
+
"loss": 0.4314,
|
| 2961 |
+
"step": 4800
|
| 2962 |
+
},
|
| 2963 |
+
{
|
| 2964 |
+
"epoch": 6.49,
|
| 2965 |
+
"learning_rate": 0.00019444444444444446,
|
| 2966 |
+
"loss": 0.405,
|
| 2967 |
+
"step": 4810
|
| 2968 |
+
},
|
| 2969 |
+
{
|
| 2970 |
+
"epoch": 6.51,
|
| 2971 |
+
"learning_rate": 0.00019369369369369368,
|
| 2972 |
+
"loss": 0.4473,
|
| 2973 |
+
"step": 4820
|
| 2974 |
+
},
|
| 2975 |
+
{
|
| 2976 |
+
"epoch": 6.52,
|
| 2977 |
+
"learning_rate": 0.00019294294294294295,
|
| 2978 |
+
"loss": 0.436,
|
| 2979 |
+
"step": 4830
|
| 2980 |
+
},
|
| 2981 |
+
{
|
| 2982 |
+
"epoch": 6.53,
|
| 2983 |
+
"learning_rate": 0.0001921921921921922,
|
| 2984 |
+
"loss": 0.4574,
|
| 2985 |
+
"step": 4840
|
| 2986 |
+
},
|
| 2987 |
+
{
|
| 2988 |
+
"epoch": 6.55,
|
| 2989 |
+
"learning_rate": 0.00019144144144144144,
|
| 2990 |
+
"loss": 0.4293,
|
| 2991 |
+
"step": 4850
|
| 2992 |
+
},
|
| 2993 |
+
{
|
| 2994 |
+
"epoch": 6.56,
|
| 2995 |
+
"learning_rate": 0.0001906906906906907,
|
| 2996 |
+
"loss": 0.3983,
|
| 2997 |
+
"step": 4860
|
| 2998 |
+
},
|
| 2999 |
+
{
|
| 3000 |
+
"epoch": 6.57,
|
| 3001 |
+
"learning_rate": 0.00018993993993993993,
|
| 3002 |
+
"loss": 0.4439,
|
| 3003 |
+
"step": 4870
|
| 3004 |
+
},
|
| 3005 |
+
{
|
| 3006 |
+
"epoch": 6.59,
|
| 3007 |
+
"learning_rate": 0.0001891891891891892,
|
| 3008 |
+
"loss": 0.4613,
|
| 3009 |
+
"step": 4880
|
| 3010 |
+
},
|
| 3011 |
+
{
|
| 3012 |
+
"epoch": 6.6,
|
| 3013 |
+
"learning_rate": 0.00018843843843843844,
|
| 3014 |
+
"loss": 0.4209,
|
| 3015 |
+
"step": 4890
|
| 3016 |
+
},
|
| 3017 |
+
{
|
| 3018 |
+
"epoch": 6.61,
|
| 3019 |
+
"learning_rate": 0.00018768768768768769,
|
| 3020 |
+
"loss": 0.5008,
|
| 3021 |
+
"step": 4900
|
| 3022 |
+
},
|
| 3023 |
+
{
|
| 3024 |
+
"epoch": 6.63,
|
| 3025 |
+
"learning_rate": 0.00018693693693693696,
|
| 3026 |
+
"loss": 0.4341,
|
| 3027 |
+
"step": 4910
|
| 3028 |
+
},
|
| 3029 |
+
{
|
| 3030 |
+
"epoch": 6.64,
|
| 3031 |
+
"learning_rate": 0.00018618618618618617,
|
| 3032 |
+
"loss": 0.4438,
|
| 3033 |
+
"step": 4920
|
| 3034 |
+
},
|
| 3035 |
+
{
|
| 3036 |
+
"epoch": 6.66,
|
| 3037 |
+
"learning_rate": 0.00018543543543543545,
|
| 3038 |
+
"loss": 0.3922,
|
| 3039 |
+
"step": 4930
|
| 3040 |
+
},
|
| 3041 |
+
{
|
| 3042 |
+
"epoch": 6.67,
|
| 3043 |
+
"learning_rate": 0.0001846846846846847,
|
| 3044 |
+
"loss": 0.4661,
|
| 3045 |
+
"step": 4940
|
| 3046 |
+
},
|
| 3047 |
+
{
|
| 3048 |
+
"epoch": 6.68,
|
| 3049 |
+
"learning_rate": 0.00018393393393393393,
|
| 3050 |
+
"loss": 0.4535,
|
| 3051 |
+
"step": 4950
|
| 3052 |
+
},
|
| 3053 |
+
{
|
| 3054 |
+
"epoch": 6.7,
|
| 3055 |
+
"learning_rate": 0.0001831831831831832,
|
| 3056 |
+
"loss": 0.4534,
|
| 3057 |
+
"step": 4960
|
| 3058 |
+
},
|
| 3059 |
+
{
|
| 3060 |
+
"epoch": 6.71,
|
| 3061 |
+
"learning_rate": 0.00018243243243243242,
|
| 3062 |
+
"loss": 0.4782,
|
| 3063 |
+
"step": 4970
|
| 3064 |
+
},
|
| 3065 |
+
{
|
| 3066 |
+
"epoch": 6.72,
|
| 3067 |
+
"learning_rate": 0.0001816816816816817,
|
| 3068 |
+
"loss": 0.4182,
|
| 3069 |
+
"step": 4980
|
| 3070 |
+
},
|
| 3071 |
+
{
|
| 3072 |
+
"epoch": 6.74,
|
| 3073 |
+
"learning_rate": 0.00018093093093093094,
|
| 3074 |
+
"loss": 0.4373,
|
| 3075 |
+
"step": 4990
|
| 3076 |
+
},
|
| 3077 |
+
{
|
| 3078 |
+
"epoch": 6.75,
|
| 3079 |
+
"learning_rate": 0.00018018018018018018,
|
| 3080 |
+
"loss": 0.4867,
|
| 3081 |
+
"step": 5000
|
| 3082 |
+
},
|
| 3083 |
+
{
|
| 3084 |
+
"epoch": 6.76,
|
| 3085 |
+
"learning_rate": 0.00017942942942942945,
|
| 3086 |
+
"loss": 0.4659,
|
| 3087 |
+
"step": 5010
|
| 3088 |
+
},
|
| 3089 |
+
{
|
| 3090 |
+
"epoch": 6.78,
|
| 3091 |
+
"learning_rate": 0.00017867867867867867,
|
| 3092 |
+
"loss": 0.3941,
|
| 3093 |
+
"step": 5020
|
| 3094 |
+
},
|
| 3095 |
+
{
|
| 3096 |
+
"epoch": 6.79,
|
| 3097 |
+
"learning_rate": 0.00017792792792792792,
|
| 3098 |
+
"loss": 0.4212,
|
| 3099 |
+
"step": 5030
|
| 3100 |
+
},
|
| 3101 |
+
{
|
| 3102 |
+
"epoch": 6.8,
|
| 3103 |
+
"learning_rate": 0.0001771771771771772,
|
| 3104 |
+
"loss": 0.4637,
|
| 3105 |
+
"step": 5040
|
| 3106 |
+
},
|
| 3107 |
+
{
|
| 3108 |
+
"epoch": 6.82,
|
| 3109 |
+
"learning_rate": 0.00017642642642642643,
|
| 3110 |
+
"loss": 0.5005,
|
| 3111 |
+
"step": 5050
|
| 3112 |
+
},
|
| 3113 |
+
{
|
| 3114 |
+
"epoch": 6.83,
|
| 3115 |
+
"learning_rate": 0.00017567567567567568,
|
| 3116 |
+
"loss": 0.4785,
|
| 3117 |
+
"step": 5060
|
| 3118 |
+
},
|
| 3119 |
+
{
|
| 3120 |
+
"epoch": 6.84,
|
| 3121 |
+
"learning_rate": 0.00017492492492492492,
|
| 3122 |
+
"loss": 0.4328,
|
| 3123 |
+
"step": 5070
|
| 3124 |
+
},
|
| 3125 |
+
{
|
| 3126 |
+
"epoch": 6.86,
|
| 3127 |
+
"learning_rate": 0.00017417417417417416,
|
| 3128 |
+
"loss": 0.3952,
|
| 3129 |
+
"step": 5080
|
| 3130 |
+
},
|
| 3131 |
+
{
|
| 3132 |
+
"epoch": 6.87,
|
| 3133 |
+
"learning_rate": 0.00017342342342342344,
|
| 3134 |
+
"loss": 0.4419,
|
| 3135 |
+
"step": 5090
|
| 3136 |
+
},
|
| 3137 |
+
{
|
| 3138 |
+
"epoch": 6.88,
|
| 3139 |
+
"learning_rate": 0.00017267267267267268,
|
| 3140 |
+
"loss": 0.4173,
|
| 3141 |
+
"step": 5100
|
| 3142 |
+
},
|
| 3143 |
+
{
|
| 3144 |
+
"epoch": 6.9,
|
| 3145 |
+
"learning_rate": 0.00017192192192192192,
|
| 3146 |
+
"loss": 0.4701,
|
| 3147 |
+
"step": 5110
|
| 3148 |
+
},
|
| 3149 |
+
{
|
| 3150 |
+
"epoch": 6.91,
|
| 3151 |
+
"learning_rate": 0.00017117117117117117,
|
| 3152 |
+
"loss": 0.4804,
|
| 3153 |
+
"step": 5120
|
| 3154 |
+
},
|
| 3155 |
+
{
|
| 3156 |
+
"epoch": 6.93,
|
| 3157 |
+
"learning_rate": 0.0001704204204204204,
|
| 3158 |
+
"loss": 0.3814,
|
| 3159 |
+
"step": 5130
|
| 3160 |
+
},
|
| 3161 |
+
{
|
| 3162 |
+
"epoch": 6.94,
|
| 3163 |
+
"learning_rate": 0.00016966966966966968,
|
| 3164 |
+
"loss": 0.4407,
|
| 3165 |
+
"step": 5140
|
| 3166 |
+
},
|
| 3167 |
+
{
|
| 3168 |
+
"epoch": 6.95,
|
| 3169 |
+
"learning_rate": 0.00016891891891891893,
|
| 3170 |
+
"loss": 0.4275,
|
| 3171 |
+
"step": 5150
|
| 3172 |
+
},
|
| 3173 |
+
{
|
| 3174 |
+
"epoch": 6.97,
|
| 3175 |
+
"learning_rate": 0.00016816816816816817,
|
| 3176 |
+
"loss": 0.4784,
|
| 3177 |
+
"step": 5160
|
| 3178 |
+
},
|
| 3179 |
+
{
|
| 3180 |
+
"epoch": 6.98,
|
| 3181 |
+
"learning_rate": 0.00016741741741741742,
|
| 3182 |
+
"loss": 0.5154,
|
| 3183 |
+
"step": 5170
|
| 3184 |
+
},
|
| 3185 |
+
{
|
| 3186 |
+
"epoch": 6.99,
|
| 3187 |
+
"learning_rate": 0.00016666666666666666,
|
| 3188 |
+
"loss": 0.4628,
|
| 3189 |
+
"step": 5180
|
| 3190 |
+
},
|
| 3191 |
+
{
|
| 3192 |
+
"epoch": 7.0,
|
| 3193 |
+
"eval_accuracy": 0.9037807561512302,
|
| 3194 |
+
"eval_f1": 0.9033003135892885,
|
| 3195 |
+
"eval_loss": 0.3493529260158539,
|
| 3196 |
+
"eval_precision": 0.9058549926677517,
|
| 3197 |
+
"eval_recall": 0.9033230091030984,
|
| 3198 |
+
"eval_runtime": 191.6635,
|
| 3199 |
+
"eval_samples_per_second": 52.164,
|
| 3200 |
+
"eval_steps_per_second": 1.936,
|
| 3201 |
+
"step": 5185
|
| 3202 |
+
},
|
| 3203 |
+
{
|
| 3204 |
+
"epoch": 7.01,
|
| 3205 |
+
"learning_rate": 0.00016591591591591593,
|
| 3206 |
+
"loss": 0.3719,
|
| 3207 |
+
"step": 5190
|
| 3208 |
+
},
|
| 3209 |
+
{
|
| 3210 |
+
"epoch": 7.02,
|
| 3211 |
+
"learning_rate": 0.00016516516516516518,
|
| 3212 |
+
"loss": 0.3562,
|
| 3213 |
+
"step": 5200
|
| 3214 |
+
},
|
| 3215 |
+
{
|
| 3216 |
+
"epoch": 7.03,
|
| 3217 |
+
"learning_rate": 0.00016441441441441442,
|
| 3218 |
+
"loss": 0.4041,
|
| 3219 |
+
"step": 5210
|
| 3220 |
+
},
|
| 3221 |
+
{
|
| 3222 |
+
"epoch": 7.05,
|
| 3223 |
+
"learning_rate": 0.00016366366366366367,
|
| 3224 |
+
"loss": 0.3744,
|
| 3225 |
+
"step": 5220
|
| 3226 |
+
},
|
| 3227 |
+
{
|
| 3228 |
+
"epoch": 7.06,
|
| 3229 |
+
"learning_rate": 0.0001629129129129129,
|
| 3230 |
+
"loss": 0.4728,
|
| 3231 |
+
"step": 5230
|
| 3232 |
+
},
|
| 3233 |
+
{
|
| 3234 |
+
"epoch": 7.07,
|
| 3235 |
+
"learning_rate": 0.00016216216216216218,
|
| 3236 |
+
"loss": 0.4138,
|
| 3237 |
+
"step": 5240
|
| 3238 |
+
},
|
| 3239 |
+
{
|
| 3240 |
+
"epoch": 7.09,
|
| 3241 |
+
"learning_rate": 0.00016141141141141143,
|
| 3242 |
+
"loss": 0.3801,
|
| 3243 |
+
"step": 5250
|
| 3244 |
+
},
|
| 3245 |
+
{
|
| 3246 |
+
"epoch": 7.1,
|
| 3247 |
+
"learning_rate": 0.00016066066066066064,
|
| 3248 |
+
"loss": 0.4113,
|
| 3249 |
+
"step": 5260
|
| 3250 |
+
},
|
| 3251 |
+
{
|
| 3252 |
+
"epoch": 7.11,
|
| 3253 |
+
"learning_rate": 0.00015990990990990991,
|
| 3254 |
+
"loss": 0.388,
|
| 3255 |
+
"step": 5270
|
| 3256 |
+
},
|
| 3257 |
+
{
|
| 3258 |
+
"epoch": 7.13,
|
| 3259 |
+
"learning_rate": 0.00015915915915915916,
|
| 3260 |
+
"loss": 0.4006,
|
| 3261 |
+
"step": 5280
|
| 3262 |
+
},
|
| 3263 |
+
{
|
| 3264 |
+
"epoch": 7.14,
|
| 3265 |
+
"learning_rate": 0.00015840840840840843,
|
| 3266 |
+
"loss": 0.3767,
|
| 3267 |
+
"step": 5290
|
| 3268 |
+
},
|
| 3269 |
+
{
|
| 3270 |
+
"epoch": 7.15,
|
| 3271 |
+
"learning_rate": 0.00015765765765765767,
|
| 3272 |
+
"loss": 0.419,
|
| 3273 |
+
"step": 5300
|
| 3274 |
+
},
|
| 3275 |
+
{
|
| 3276 |
+
"epoch": 7.17,
|
| 3277 |
+
"learning_rate": 0.0001569069069069069,
|
| 3278 |
+
"loss": 0.3563,
|
| 3279 |
+
"step": 5310
|
| 3280 |
+
},
|
| 3281 |
+
{
|
| 3282 |
+
"epoch": 7.18,
|
| 3283 |
+
"learning_rate": 0.00015615615615615616,
|
| 3284 |
+
"loss": 0.4191,
|
| 3285 |
+
"step": 5320
|
| 3286 |
+
},
|
| 3287 |
+
{
|
| 3288 |
+
"epoch": 7.2,
|
| 3289 |
+
"learning_rate": 0.0001554054054054054,
|
| 3290 |
+
"loss": 0.4491,
|
| 3291 |
+
"step": 5330
|
| 3292 |
+
},
|
| 3293 |
+
{
|
| 3294 |
+
"epoch": 7.21,
|
| 3295 |
+
"learning_rate": 0.00015465465465465468,
|
| 3296 |
+
"loss": 0.3764,
|
| 3297 |
+
"step": 5340
|
| 3298 |
+
},
|
| 3299 |
+
{
|
| 3300 |
+
"epoch": 7.22,
|
| 3301 |
+
"learning_rate": 0.0001539039039039039,
|
| 3302 |
+
"loss": 0.3798,
|
| 3303 |
+
"step": 5350
|
| 3304 |
+
},
|
| 3305 |
+
{
|
| 3306 |
+
"epoch": 7.24,
|
| 3307 |
+
"learning_rate": 0.00015315315315315314,
|
| 3308 |
+
"loss": 0.4415,
|
| 3309 |
+
"step": 5360
|
| 3310 |
+
},
|
| 3311 |
+
{
|
| 3312 |
+
"epoch": 7.25,
|
| 3313 |
+
"learning_rate": 0.0001524024024024024,
|
| 3314 |
+
"loss": 0.3818,
|
| 3315 |
+
"step": 5370
|
| 3316 |
+
},
|
| 3317 |
+
{
|
| 3318 |
+
"epoch": 7.26,
|
| 3319 |
+
"learning_rate": 0.00015165165165165165,
|
| 3320 |
+
"loss": 0.3974,
|
| 3321 |
+
"step": 5380
|
| 3322 |
+
},
|
| 3323 |
+
{
|
| 3324 |
+
"epoch": 7.28,
|
| 3325 |
+
"learning_rate": 0.00015090090090090093,
|
| 3326 |
+
"loss": 0.4371,
|
| 3327 |
+
"step": 5390
|
| 3328 |
+
},
|
| 3329 |
+
{
|
| 3330 |
+
"epoch": 7.29,
|
| 3331 |
+
"learning_rate": 0.00015015015015015014,
|
| 3332 |
+
"loss": 0.4014,
|
| 3333 |
+
"step": 5400
|
| 3334 |
+
},
|
| 3335 |
+
{
|
| 3336 |
+
"epoch": 7.3,
|
| 3337 |
+
"learning_rate": 0.0001493993993993994,
|
| 3338 |
+
"loss": 0.3844,
|
| 3339 |
+
"step": 5410
|
| 3340 |
+
},
|
| 3341 |
+
{
|
| 3342 |
+
"epoch": 7.32,
|
| 3343 |
+
"learning_rate": 0.00014864864864864866,
|
| 3344 |
+
"loss": 0.384,
|
| 3345 |
+
"step": 5420
|
| 3346 |
+
},
|
| 3347 |
+
{
|
| 3348 |
+
"epoch": 7.33,
|
| 3349 |
+
"learning_rate": 0.0001478978978978979,
|
| 3350 |
+
"loss": 0.4619,
|
| 3351 |
+
"step": 5430
|
| 3352 |
+
},
|
| 3353 |
+
{
|
| 3354 |
+
"epoch": 7.34,
|
| 3355 |
+
"learning_rate": 0.00014714714714714715,
|
| 3356 |
+
"loss": 0.3962,
|
| 3357 |
+
"step": 5440
|
| 3358 |
+
},
|
| 3359 |
+
{
|
| 3360 |
+
"epoch": 7.36,
|
| 3361 |
+
"learning_rate": 0.0001463963963963964,
|
| 3362 |
+
"loss": 0.4337,
|
| 3363 |
+
"step": 5450
|
| 3364 |
+
},
|
| 3365 |
+
{
|
| 3366 |
+
"epoch": 7.37,
|
| 3367 |
+
"learning_rate": 0.00014564564564564564,
|
| 3368 |
+
"loss": 0.4088,
|
| 3369 |
+
"step": 5460
|
| 3370 |
+
},
|
| 3371 |
+
{
|
| 3372 |
+
"epoch": 7.38,
|
| 3373 |
+
"learning_rate": 0.0001448948948948949,
|
| 3374 |
+
"loss": 0.3786,
|
| 3375 |
+
"step": 5470
|
| 3376 |
+
},
|
| 3377 |
+
{
|
| 3378 |
+
"epoch": 7.4,
|
| 3379 |
+
"learning_rate": 0.00014414414414414415,
|
| 3380 |
+
"loss": 0.4226,
|
| 3381 |
+
"step": 5480
|
| 3382 |
+
},
|
| 3383 |
+
{
|
| 3384 |
+
"epoch": 7.41,
|
| 3385 |
+
"learning_rate": 0.0001433933933933934,
|
| 3386 |
+
"loss": 0.44,
|
| 3387 |
+
"step": 5490
|
| 3388 |
+
},
|
| 3389 |
+
{
|
| 3390 |
+
"epoch": 7.42,
|
| 3391 |
+
"learning_rate": 0.00014264264264264264,
|
| 3392 |
+
"loss": 0.4322,
|
| 3393 |
+
"step": 5500
|
| 3394 |
+
},
|
| 3395 |
+
{
|
| 3396 |
+
"epoch": 7.44,
|
| 3397 |
+
"learning_rate": 0.00014189189189189188,
|
| 3398 |
+
"loss": 0.3269,
|
| 3399 |
+
"step": 5510
|
| 3400 |
+
},
|
| 3401 |
+
{
|
| 3402 |
+
"epoch": 7.45,
|
| 3403 |
+
"learning_rate": 0.00014114114114114116,
|
| 3404 |
+
"loss": 0.347,
|
| 3405 |
+
"step": 5520
|
| 3406 |
+
},
|
| 3407 |
+
{
|
| 3408 |
+
"epoch": 7.47,
|
| 3409 |
+
"learning_rate": 0.0001403903903903904,
|
| 3410 |
+
"loss": 0.4496,
|
| 3411 |
+
"step": 5530
|
| 3412 |
+
},
|
| 3413 |
+
{
|
| 3414 |
+
"epoch": 7.48,
|
| 3415 |
+
"learning_rate": 0.00013963963963963964,
|
| 3416 |
+
"loss": 0.4092,
|
| 3417 |
+
"step": 5540
|
| 3418 |
+
},
|
| 3419 |
+
{
|
| 3420 |
+
"epoch": 7.49,
|
| 3421 |
+
"learning_rate": 0.0001388888888888889,
|
| 3422 |
+
"loss": 0.3484,
|
| 3423 |
+
"step": 5550
|
| 3424 |
+
},
|
| 3425 |
+
{
|
| 3426 |
+
"epoch": 7.51,
|
| 3427 |
+
"learning_rate": 0.00013813813813813813,
|
| 3428 |
+
"loss": 0.4184,
|
| 3429 |
+
"step": 5560
|
| 3430 |
+
},
|
| 3431 |
+
{
|
| 3432 |
+
"epoch": 7.52,
|
| 3433 |
+
"learning_rate": 0.0001373873873873874,
|
| 3434 |
+
"loss": 0.3967,
|
| 3435 |
+
"step": 5570
|
| 3436 |
+
},
|
| 3437 |
+
{
|
| 3438 |
+
"epoch": 7.53,
|
| 3439 |
+
"learning_rate": 0.00013663663663663665,
|
| 3440 |
+
"loss": 0.4125,
|
| 3441 |
+
"step": 5580
|
| 3442 |
+
},
|
| 3443 |
+
{
|
| 3444 |
+
"epoch": 7.55,
|
| 3445 |
+
"learning_rate": 0.00013588588588588587,
|
| 3446 |
+
"loss": 0.401,
|
| 3447 |
+
"step": 5590
|
| 3448 |
+
},
|
| 3449 |
+
{
|
| 3450 |
+
"epoch": 7.56,
|
| 3451 |
+
"learning_rate": 0.00013513513513513514,
|
| 3452 |
+
"loss": 0.4787,
|
| 3453 |
+
"step": 5600
|
| 3454 |
+
},
|
| 3455 |
+
{
|
| 3456 |
+
"epoch": 7.57,
|
| 3457 |
+
"learning_rate": 0.00013438438438438438,
|
| 3458 |
+
"loss": 0.3997,
|
| 3459 |
+
"step": 5610
|
| 3460 |
+
},
|
| 3461 |
+
{
|
| 3462 |
+
"epoch": 7.59,
|
| 3463 |
+
"learning_rate": 0.00013363363363363365,
|
| 3464 |
+
"loss": 0.3598,
|
| 3465 |
+
"step": 5620
|
| 3466 |
+
},
|
| 3467 |
+
{
|
| 3468 |
+
"epoch": 7.6,
|
| 3469 |
+
"learning_rate": 0.0001328828828828829,
|
| 3470 |
+
"loss": 0.3988,
|
| 3471 |
+
"step": 5630
|
| 3472 |
+
},
|
| 3473 |
+
{
|
| 3474 |
+
"epoch": 7.61,
|
| 3475 |
+
"learning_rate": 0.00013213213213213211,
|
| 3476 |
+
"loss": 0.3209,
|
| 3477 |
+
"step": 5640
|
| 3478 |
+
},
|
| 3479 |
+
{
|
| 3480 |
+
"epoch": 7.63,
|
| 3481 |
+
"learning_rate": 0.00013138138138138139,
|
| 3482 |
+
"loss": 0.3528,
|
| 3483 |
+
"step": 5650
|
| 3484 |
+
},
|
| 3485 |
+
{
|
| 3486 |
+
"epoch": 7.64,
|
| 3487 |
+
"learning_rate": 0.00013063063063063063,
|
| 3488 |
+
"loss": 0.4024,
|
| 3489 |
+
"step": 5660
|
| 3490 |
+
},
|
| 3491 |
+
{
|
| 3492 |
+
"epoch": 7.65,
|
| 3493 |
+
"learning_rate": 0.00012987987987987987,
|
| 3494 |
+
"loss": 0.3689,
|
| 3495 |
+
"step": 5670
|
| 3496 |
+
},
|
| 3497 |
+
{
|
| 3498 |
+
"epoch": 7.67,
|
| 3499 |
+
"learning_rate": 0.00012912912912912915,
|
| 3500 |
+
"loss": 0.3908,
|
| 3501 |
+
"step": 5680
|
| 3502 |
+
},
|
| 3503 |
+
{
|
| 3504 |
+
"epoch": 7.68,
|
| 3505 |
+
"learning_rate": 0.00012837837837837836,
|
| 3506 |
+
"loss": 0.4048,
|
| 3507 |
+
"step": 5690
|
| 3508 |
+
},
|
| 3509 |
+
{
|
| 3510 |
+
"epoch": 7.69,
|
| 3511 |
+
"learning_rate": 0.00012762762762762763,
|
| 3512 |
+
"loss": 0.3461,
|
| 3513 |
+
"step": 5700
|
| 3514 |
+
},
|
| 3515 |
+
{
|
| 3516 |
+
"epoch": 7.71,
|
| 3517 |
+
"learning_rate": 0.00012687687687687688,
|
| 3518 |
+
"loss": 0.4064,
|
| 3519 |
+
"step": 5710
|
| 3520 |
+
},
|
| 3521 |
+
{
|
| 3522 |
+
"epoch": 7.72,
|
| 3523 |
+
"learning_rate": 0.00012612612612612612,
|
| 3524 |
+
"loss": 0.3859,
|
| 3525 |
+
"step": 5720
|
| 3526 |
+
},
|
| 3527 |
+
{
|
| 3528 |
+
"epoch": 7.74,
|
| 3529 |
+
"learning_rate": 0.0001253753753753754,
|
| 3530 |
+
"loss": 0.3992,
|
| 3531 |
+
"step": 5730
|
| 3532 |
+
},
|
| 3533 |
+
{
|
| 3534 |
+
"epoch": 7.75,
|
| 3535 |
+
"learning_rate": 0.0001246246246246246,
|
| 3536 |
+
"loss": 0.3116,
|
| 3537 |
+
"step": 5740
|
| 3538 |
+
},
|
| 3539 |
+
{
|
| 3540 |
+
"epoch": 7.76,
|
| 3541 |
+
"learning_rate": 0.00012387387387387388,
|
| 3542 |
+
"loss": 0.3627,
|
| 3543 |
+
"step": 5750
|
| 3544 |
+
},
|
| 3545 |
+
{
|
| 3546 |
+
"epoch": 7.78,
|
| 3547 |
+
"learning_rate": 0.00012312312312312313,
|
| 3548 |
+
"loss": 0.3946,
|
| 3549 |
+
"step": 5760
|
| 3550 |
+
},
|
| 3551 |
+
{
|
| 3552 |
+
"epoch": 7.79,
|
| 3553 |
+
"learning_rate": 0.00012237237237237237,
|
| 3554 |
+
"loss": 0.3297,
|
| 3555 |
+
"step": 5770
|
| 3556 |
+
},
|
| 3557 |
+
{
|
| 3558 |
+
"epoch": 7.8,
|
| 3559 |
+
"learning_rate": 0.00012162162162162163,
|
| 3560 |
+
"loss": 0.4199,
|
| 3561 |
+
"step": 5780
|
| 3562 |
+
},
|
| 3563 |
+
{
|
| 3564 |
+
"epoch": 7.82,
|
| 3565 |
+
"learning_rate": 0.00012087087087087087,
|
| 3566 |
+
"loss": 0.4164,
|
| 3567 |
+
"step": 5790
|
| 3568 |
+
},
|
| 3569 |
+
{
|
| 3570 |
+
"epoch": 7.83,
|
| 3571 |
+
"learning_rate": 0.00012012012012012012,
|
| 3572 |
+
"loss": 0.3763,
|
| 3573 |
+
"step": 5800
|
| 3574 |
+
},
|
| 3575 |
+
{
|
| 3576 |
+
"epoch": 7.84,
|
| 3577 |
+
"learning_rate": 0.00011936936936936938,
|
| 3578 |
+
"loss": 0.3401,
|
| 3579 |
+
"step": 5810
|
| 3580 |
+
},
|
| 3581 |
+
{
|
| 3582 |
+
"epoch": 7.86,
|
| 3583 |
+
"learning_rate": 0.00011861861861861862,
|
| 3584 |
+
"loss": 0.3524,
|
| 3585 |
+
"step": 5820
|
| 3586 |
+
},
|
| 3587 |
+
{
|
| 3588 |
+
"epoch": 7.87,
|
| 3589 |
+
"learning_rate": 0.00011786786786786786,
|
| 3590 |
+
"loss": 0.4263,
|
| 3591 |
+
"step": 5830
|
| 3592 |
+
},
|
| 3593 |
+
{
|
| 3594 |
+
"epoch": 7.88,
|
| 3595 |
+
"learning_rate": 0.00011711711711711712,
|
| 3596 |
+
"loss": 0.3855,
|
| 3597 |
+
"step": 5840
|
| 3598 |
+
},
|
| 3599 |
+
{
|
| 3600 |
+
"epoch": 7.9,
|
| 3601 |
+
"learning_rate": 0.00011636636636636637,
|
| 3602 |
+
"loss": 0.3678,
|
| 3603 |
+
"step": 5850
|
| 3604 |
+
},
|
| 3605 |
+
{
|
| 3606 |
+
"epoch": 7.91,
|
| 3607 |
+
"learning_rate": 0.00011561561561561562,
|
| 3608 |
+
"loss": 0.3067,
|
| 3609 |
+
"step": 5860
|
| 3610 |
+
},
|
| 3611 |
+
{
|
| 3612 |
+
"epoch": 7.92,
|
| 3613 |
+
"learning_rate": 0.00011486486486486487,
|
| 3614 |
+
"loss": 0.3637,
|
| 3615 |
+
"step": 5870
|
| 3616 |
+
},
|
| 3617 |
+
{
|
| 3618 |
+
"epoch": 7.94,
|
| 3619 |
+
"learning_rate": 0.00011411411411411411,
|
| 3620 |
+
"loss": 0.4032,
|
| 3621 |
+
"step": 5880
|
| 3622 |
+
},
|
| 3623 |
+
{
|
| 3624 |
+
"epoch": 7.95,
|
| 3625 |
+
"learning_rate": 0.00011336336336336337,
|
| 3626 |
+
"loss": 0.3832,
|
| 3627 |
+
"step": 5890
|
| 3628 |
+
},
|
| 3629 |
+
{
|
| 3630 |
+
"epoch": 7.96,
|
| 3631 |
+
"learning_rate": 0.00011261261261261261,
|
| 3632 |
+
"loss": 0.4019,
|
| 3633 |
+
"step": 5900
|
| 3634 |
+
},
|
| 3635 |
+
{
|
| 3636 |
+
"epoch": 7.98,
|
| 3637 |
+
"learning_rate": 0.00011186186186186187,
|
| 3638 |
+
"loss": 0.4264,
|
| 3639 |
+
"step": 5910
|
| 3640 |
+
},
|
| 3641 |
+
{
|
| 3642 |
+
"epoch": 7.99,
|
| 3643 |
+
"learning_rate": 0.0001111111111111111,
|
| 3644 |
+
"loss": 0.3256,
|
| 3645 |
+
"step": 5920
|
| 3646 |
+
},
|
| 3647 |
+
{
|
| 3648 |
+
"epoch": 8.0,
|
| 3649 |
+
"eval_accuracy": 0.9149829965993198,
|
| 3650 |
+
"eval_f1": 0.9145293036070032,
|
| 3651 |
+
"eval_loss": 0.3138176500797272,
|
| 3652 |
+
"eval_precision": 0.9170611788545995,
|
| 3653 |
+
"eval_recall": 0.9140792071407026,
|
| 3654 |
+
"eval_runtime": 191.5713,
|
| 3655 |
+
"eval_samples_per_second": 52.189,
|
| 3656 |
+
"eval_steps_per_second": 1.937,
|
| 3657 |
+
"step": 5926
|
| 3658 |
+
},
|
| 3659 |
+
{
|
| 3660 |
+
"epoch": 8.01,
|
| 3661 |
+
"learning_rate": 0.00011036036036036036,
|
| 3662 |
+
"loss": 0.3767,
|
| 3663 |
+
"step": 5930
|
| 3664 |
+
},
|
| 3665 |
+
{
|
| 3666 |
+
"epoch": 8.02,
|
| 3667 |
+
"learning_rate": 0.00010960960960960962,
|
| 3668 |
+
"loss": 0.3465,
|
| 3669 |
+
"step": 5940
|
| 3670 |
+
},
|
| 3671 |
+
{
|
| 3672 |
+
"epoch": 8.03,
|
| 3673 |
+
"learning_rate": 0.00010885885885885886,
|
| 3674 |
+
"loss": 0.3514,
|
| 3675 |
+
"step": 5950
|
| 3676 |
+
},
|
| 3677 |
+
{
|
| 3678 |
+
"epoch": 8.05,
|
| 3679 |
+
"learning_rate": 0.00010810810810810812,
|
| 3680 |
+
"loss": 0.3637,
|
| 3681 |
+
"step": 5960
|
| 3682 |
+
},
|
| 3683 |
+
{
|
| 3684 |
+
"epoch": 8.06,
|
| 3685 |
+
"learning_rate": 0.00010735735735735735,
|
| 3686 |
+
"loss": 0.3834,
|
| 3687 |
+
"step": 5970
|
| 3688 |
+
},
|
| 3689 |
+
{
|
| 3690 |
+
"epoch": 8.07,
|
| 3691 |
+
"learning_rate": 0.00010660660660660661,
|
| 3692 |
+
"loss": 0.2656,
|
| 3693 |
+
"step": 5980
|
| 3694 |
+
},
|
| 3695 |
+
{
|
| 3696 |
+
"epoch": 8.09,
|
| 3697 |
+
"learning_rate": 0.00010585585585585585,
|
| 3698 |
+
"loss": 0.3559,
|
| 3699 |
+
"step": 5990
|
| 3700 |
+
},
|
| 3701 |
+
{
|
| 3702 |
+
"epoch": 8.1,
|
| 3703 |
+
"learning_rate": 0.00010510510510510511,
|
| 3704 |
+
"loss": 0.3491,
|
| 3705 |
+
"step": 6000
|
| 3706 |
+
},
|
| 3707 |
+
{
|
| 3708 |
+
"epoch": 8.11,
|
| 3709 |
+
"learning_rate": 0.00010435435435435436,
|
| 3710 |
+
"loss": 0.4012,
|
| 3711 |
+
"step": 6010
|
| 3712 |
+
},
|
| 3713 |
+
{
|
| 3714 |
+
"epoch": 8.13,
|
| 3715 |
+
"learning_rate": 0.0001036036036036036,
|
| 3716 |
+
"loss": 0.3757,
|
| 3717 |
+
"step": 6020
|
| 3718 |
+
},
|
| 3719 |
+
{
|
| 3720 |
+
"epoch": 8.14,
|
| 3721 |
+
"learning_rate": 0.00010285285285285286,
|
| 3722 |
+
"loss": 0.3678,
|
| 3723 |
+
"step": 6030
|
| 3724 |
+
},
|
| 3725 |
+
{
|
| 3726 |
+
"epoch": 8.15,
|
| 3727 |
+
"learning_rate": 0.0001021021021021021,
|
| 3728 |
+
"loss": 0.3667,
|
| 3729 |
+
"step": 6040
|
| 3730 |
+
},
|
| 3731 |
+
{
|
| 3732 |
+
"epoch": 8.17,
|
| 3733 |
+
"learning_rate": 0.00010135135135135136,
|
| 3734 |
+
"loss": 0.3421,
|
| 3735 |
+
"step": 6050
|
| 3736 |
+
},
|
| 3737 |
+
{
|
| 3738 |
+
"epoch": 8.18,
|
| 3739 |
+
"learning_rate": 0.0001006006006006006,
|
| 3740 |
+
"loss": 0.3924,
|
| 3741 |
+
"step": 6060
|
| 3742 |
+
},
|
| 3743 |
+
{
|
| 3744 |
+
"epoch": 8.19,
|
| 3745 |
+
"learning_rate": 9.984984984984985e-05,
|
| 3746 |
+
"loss": 0.2621,
|
| 3747 |
+
"step": 6070
|
| 3748 |
+
},
|
| 3749 |
+
{
|
| 3750 |
+
"epoch": 8.21,
|
| 3751 |
+
"learning_rate": 9.90990990990991e-05,
|
| 3752 |
+
"loss": 0.3959,
|
| 3753 |
+
"step": 6080
|
| 3754 |
+
},
|
| 3755 |
+
{
|
| 3756 |
+
"epoch": 8.22,
|
| 3757 |
+
"learning_rate": 9.834834834834835e-05,
|
| 3758 |
+
"loss": 0.3191,
|
| 3759 |
+
"step": 6090
|
| 3760 |
+
},
|
| 3761 |
+
{
|
| 3762 |
+
"epoch": 8.23,
|
| 3763 |
+
"learning_rate": 9.75975975975976e-05,
|
| 3764 |
+
"loss": 0.3449,
|
| 3765 |
+
"step": 6100
|
| 3766 |
+
},
|
| 3767 |
+
{
|
| 3768 |
+
"epoch": 8.25,
|
| 3769 |
+
"learning_rate": 9.684684684684684e-05,
|
| 3770 |
+
"loss": 0.3142,
|
| 3771 |
+
"step": 6110
|
| 3772 |
+
},
|
| 3773 |
+
{
|
| 3774 |
+
"epoch": 8.26,
|
| 3775 |
+
"learning_rate": 9.60960960960961e-05,
|
| 3776 |
+
"loss": 0.3093,
|
| 3777 |
+
"step": 6120
|
| 3778 |
+
},
|
| 3779 |
+
{
|
| 3780 |
+
"epoch": 8.28,
|
| 3781 |
+
"learning_rate": 9.534534534534535e-05,
|
| 3782 |
+
"loss": 0.3446,
|
| 3783 |
+
"step": 6130
|
| 3784 |
+
},
|
| 3785 |
+
{
|
| 3786 |
+
"epoch": 8.29,
|
| 3787 |
+
"learning_rate": 9.45945945945946e-05,
|
| 3788 |
+
"loss": 0.2899,
|
| 3789 |
+
"step": 6140
|
| 3790 |
+
},
|
| 3791 |
+
{
|
| 3792 |
+
"epoch": 8.3,
|
| 3793 |
+
"learning_rate": 9.384384384384384e-05,
|
| 3794 |
+
"loss": 0.3068,
|
| 3795 |
+
"step": 6150
|
| 3796 |
+
},
|
| 3797 |
+
{
|
| 3798 |
+
"epoch": 8.32,
|
| 3799 |
+
"learning_rate": 9.309309309309309e-05,
|
| 3800 |
+
"loss": 0.3611,
|
| 3801 |
+
"step": 6160
|
| 3802 |
+
},
|
| 3803 |
+
{
|
| 3804 |
+
"epoch": 8.33,
|
| 3805 |
+
"learning_rate": 9.234234234234235e-05,
|
| 3806 |
+
"loss": 0.2767,
|
| 3807 |
+
"step": 6170
|
| 3808 |
+
},
|
| 3809 |
+
{
|
| 3810 |
+
"epoch": 8.34,
|
| 3811 |
+
"learning_rate": 9.15915915915916e-05,
|
| 3812 |
+
"loss": 0.4126,
|
| 3813 |
+
"step": 6180
|
| 3814 |
+
},
|
| 3815 |
+
{
|
| 3816 |
+
"epoch": 8.36,
|
| 3817 |
+
"learning_rate": 9.084084084084085e-05,
|
| 3818 |
+
"loss": 0.3741,
|
| 3819 |
+
"step": 6190
|
| 3820 |
+
},
|
| 3821 |
+
{
|
| 3822 |
+
"epoch": 8.37,
|
| 3823 |
+
"learning_rate": 9.009009009009009e-05,
|
| 3824 |
+
"loss": 0.3326,
|
| 3825 |
+
"step": 6200
|
| 3826 |
+
},
|
| 3827 |
+
{
|
| 3828 |
+
"epoch": 8.38,
|
| 3829 |
+
"learning_rate": 8.933933933933934e-05,
|
| 3830 |
+
"loss": 0.3034,
|
| 3831 |
+
"step": 6210
|
| 3832 |
+
},
|
| 3833 |
+
{
|
| 3834 |
+
"epoch": 8.4,
|
| 3835 |
+
"learning_rate": 8.85885885885886e-05,
|
| 3836 |
+
"loss": 0.3002,
|
| 3837 |
+
"step": 6220
|
| 3838 |
+
},
|
| 3839 |
+
{
|
| 3840 |
+
"epoch": 8.41,
|
| 3841 |
+
"learning_rate": 8.783783783783784e-05,
|
| 3842 |
+
"loss": 0.3728,
|
| 3843 |
+
"step": 6230
|
| 3844 |
+
},
|
| 3845 |
+
{
|
| 3846 |
+
"epoch": 8.42,
|
| 3847 |
+
"learning_rate": 8.708708708708708e-05,
|
| 3848 |
+
"loss": 0.3149,
|
| 3849 |
+
"step": 6240
|
| 3850 |
+
},
|
| 3851 |
+
{
|
| 3852 |
+
"epoch": 8.44,
|
| 3853 |
+
"learning_rate": 8.633633633633634e-05,
|
| 3854 |
+
"loss": 0.3122,
|
| 3855 |
+
"step": 6250
|
| 3856 |
+
},
|
| 3857 |
+
{
|
| 3858 |
+
"epoch": 8.45,
|
| 3859 |
+
"learning_rate": 8.558558558558558e-05,
|
| 3860 |
+
"loss": 0.3404,
|
| 3861 |
+
"step": 6260
|
| 3862 |
+
},
|
| 3863 |
+
{
|
| 3864 |
+
"epoch": 8.46,
|
| 3865 |
+
"learning_rate": 8.483483483483484e-05,
|
| 3866 |
+
"loss": 0.3521,
|
| 3867 |
+
"step": 6270
|
| 3868 |
+
},
|
| 3869 |
+
{
|
| 3870 |
+
"epoch": 8.48,
|
| 3871 |
+
"learning_rate": 8.408408408408409e-05,
|
| 3872 |
+
"loss": 0.3154,
|
| 3873 |
+
"step": 6280
|
| 3874 |
+
},
|
| 3875 |
+
{
|
| 3876 |
+
"epoch": 8.49,
|
| 3877 |
+
"learning_rate": 8.333333333333333e-05,
|
| 3878 |
+
"loss": 0.3654,
|
| 3879 |
+
"step": 6290
|
| 3880 |
+
},
|
| 3881 |
+
{
|
| 3882 |
+
"epoch": 8.5,
|
| 3883 |
+
"learning_rate": 8.258258258258259e-05,
|
| 3884 |
+
"loss": 0.3591,
|
| 3885 |
+
"step": 6300
|
| 3886 |
+
},
|
| 3887 |
+
{
|
| 3888 |
+
"epoch": 8.52,
|
| 3889 |
+
"learning_rate": 8.183183183183183e-05,
|
| 3890 |
+
"loss": 0.3284,
|
| 3891 |
+
"step": 6310
|
| 3892 |
+
},
|
| 3893 |
+
{
|
| 3894 |
+
"epoch": 8.53,
|
| 3895 |
+
"learning_rate": 8.108108108108109e-05,
|
| 3896 |
+
"loss": 0.2774,
|
| 3897 |
+
"step": 6320
|
| 3898 |
+
},
|
| 3899 |
+
{
|
| 3900 |
+
"epoch": 8.55,
|
| 3901 |
+
"learning_rate": 8.033033033033032e-05,
|
| 3902 |
+
"loss": 0.2813,
|
| 3903 |
+
"step": 6330
|
| 3904 |
+
},
|
| 3905 |
+
{
|
| 3906 |
+
"epoch": 8.56,
|
| 3907 |
+
"learning_rate": 7.957957957957958e-05,
|
| 3908 |
+
"loss": 0.3347,
|
| 3909 |
+
"step": 6340
|
| 3910 |
+
},
|
| 3911 |
+
{
|
| 3912 |
+
"epoch": 8.57,
|
| 3913 |
+
"learning_rate": 7.882882882882884e-05,
|
| 3914 |
+
"loss": 0.2895,
|
| 3915 |
+
"step": 6350
|
| 3916 |
+
},
|
| 3917 |
+
{
|
| 3918 |
+
"epoch": 8.59,
|
| 3919 |
+
"learning_rate": 7.807807807807808e-05,
|
| 3920 |
+
"loss": 0.3634,
|
| 3921 |
+
"step": 6360
|
| 3922 |
+
},
|
| 3923 |
+
{
|
| 3924 |
+
"epoch": 8.6,
|
| 3925 |
+
"learning_rate": 7.732732732732734e-05,
|
| 3926 |
+
"loss": 0.3155,
|
| 3927 |
+
"step": 6370
|
| 3928 |
+
},
|
| 3929 |
+
{
|
| 3930 |
+
"epoch": 8.61,
|
| 3931 |
+
"learning_rate": 7.657657657657657e-05,
|
| 3932 |
+
"loss": 0.3407,
|
| 3933 |
+
"step": 6380
|
| 3934 |
+
},
|
| 3935 |
+
{
|
| 3936 |
+
"epoch": 8.63,
|
| 3937 |
+
"learning_rate": 7.582582582582583e-05,
|
| 3938 |
+
"loss": 0.318,
|
| 3939 |
+
"step": 6390
|
| 3940 |
+
},
|
| 3941 |
+
{
|
| 3942 |
+
"epoch": 8.64,
|
| 3943 |
+
"learning_rate": 7.507507507507507e-05,
|
| 3944 |
+
"loss": 0.2963,
|
| 3945 |
+
"step": 6400
|
| 3946 |
+
},
|
| 3947 |
+
{
|
| 3948 |
+
"epoch": 8.65,
|
| 3949 |
+
"learning_rate": 7.432432432432433e-05,
|
| 3950 |
+
"loss": 0.3049,
|
| 3951 |
+
"step": 6410
|
| 3952 |
+
},
|
| 3953 |
+
{
|
| 3954 |
+
"epoch": 8.67,
|
| 3955 |
+
"learning_rate": 7.357357357357357e-05,
|
| 3956 |
+
"loss": 0.3612,
|
| 3957 |
+
"step": 6420
|
| 3958 |
+
},
|
| 3959 |
+
{
|
| 3960 |
+
"epoch": 8.68,
|
| 3961 |
+
"learning_rate": 7.282282282282282e-05,
|
| 3962 |
+
"loss": 0.3931,
|
| 3963 |
+
"step": 6430
|
| 3964 |
+
},
|
| 3965 |
+
{
|
| 3966 |
+
"epoch": 8.69,
|
| 3967 |
+
"learning_rate": 7.207207207207208e-05,
|
| 3968 |
+
"loss": 0.3213,
|
| 3969 |
+
"step": 6440
|
| 3970 |
+
},
|
| 3971 |
+
{
|
| 3972 |
+
"epoch": 8.71,
|
| 3973 |
+
"learning_rate": 7.132132132132132e-05,
|
| 3974 |
+
"loss": 0.3216,
|
| 3975 |
+
"step": 6450
|
| 3976 |
+
},
|
| 3977 |
+
{
|
| 3978 |
+
"epoch": 8.72,
|
| 3979 |
+
"learning_rate": 7.057057057057058e-05,
|
| 3980 |
+
"loss": 0.341,
|
| 3981 |
+
"step": 6460
|
| 3982 |
+
},
|
| 3983 |
+
{
|
| 3984 |
+
"epoch": 8.73,
|
| 3985 |
+
"learning_rate": 6.981981981981982e-05,
|
| 3986 |
+
"loss": 0.2771,
|
| 3987 |
+
"step": 6470
|
| 3988 |
+
},
|
| 3989 |
+
{
|
| 3990 |
+
"epoch": 8.75,
|
| 3991 |
+
"learning_rate": 6.906906906906907e-05,
|
| 3992 |
+
"loss": 0.3066,
|
| 3993 |
+
"step": 6480
|
| 3994 |
+
},
|
| 3995 |
+
{
|
| 3996 |
+
"epoch": 8.76,
|
| 3997 |
+
"learning_rate": 6.831831831831832e-05,
|
| 3998 |
+
"loss": 0.2938,
|
| 3999 |
+
"step": 6490
|
| 4000 |
+
},
|
| 4001 |
+
{
|
| 4002 |
+
"epoch": 8.77,
|
| 4003 |
+
"learning_rate": 6.756756756756757e-05,
|
| 4004 |
+
"loss": 0.3225,
|
| 4005 |
+
"step": 6500
|
| 4006 |
+
},
|
| 4007 |
+
{
|
| 4008 |
+
"epoch": 8.79,
|
| 4009 |
+
"learning_rate": 6.681681681681683e-05,
|
| 4010 |
+
"loss": 0.2781,
|
| 4011 |
+
"step": 6510
|
| 4012 |
+
},
|
| 4013 |
+
{
|
| 4014 |
+
"epoch": 8.8,
|
| 4015 |
+
"learning_rate": 6.606606606606606e-05,
|
| 4016 |
+
"loss": 0.3331,
|
| 4017 |
+
"step": 6520
|
| 4018 |
+
},
|
| 4019 |
+
{
|
| 4020 |
+
"epoch": 8.82,
|
| 4021 |
+
"learning_rate": 6.531531531531531e-05,
|
| 4022 |
+
"loss": 0.3598,
|
| 4023 |
+
"step": 6530
|
| 4024 |
+
},
|
| 4025 |
+
{
|
| 4026 |
+
"epoch": 8.83,
|
| 4027 |
+
"learning_rate": 6.456456456456457e-05,
|
| 4028 |
+
"loss": 0.325,
|
| 4029 |
+
"step": 6540
|
| 4030 |
+
},
|
| 4031 |
+
{
|
| 4032 |
+
"epoch": 8.84,
|
| 4033 |
+
"learning_rate": 6.381381381381382e-05,
|
| 4034 |
+
"loss": 0.3053,
|
| 4035 |
+
"step": 6550
|
| 4036 |
+
},
|
| 4037 |
+
{
|
| 4038 |
+
"epoch": 8.86,
|
| 4039 |
+
"learning_rate": 6.306306306306306e-05,
|
| 4040 |
+
"loss": 0.2996,
|
| 4041 |
+
"step": 6560
|
| 4042 |
+
},
|
| 4043 |
+
{
|
| 4044 |
+
"epoch": 8.87,
|
| 4045 |
+
"learning_rate": 6.23123123123123e-05,
|
| 4046 |
+
"loss": 0.3505,
|
| 4047 |
+
"step": 6570
|
| 4048 |
+
},
|
| 4049 |
+
{
|
| 4050 |
+
"epoch": 8.88,
|
| 4051 |
+
"learning_rate": 6.156156156156156e-05,
|
| 4052 |
+
"loss": 0.2732,
|
| 4053 |
+
"step": 6580
|
| 4054 |
+
},
|
| 4055 |
+
{
|
| 4056 |
+
"epoch": 8.9,
|
| 4057 |
+
"learning_rate": 6.0810810810810814e-05,
|
| 4058 |
+
"loss": 0.3205,
|
| 4059 |
+
"step": 6590
|
| 4060 |
+
},
|
| 4061 |
+
{
|
| 4062 |
+
"epoch": 8.91,
|
| 4063 |
+
"learning_rate": 6.006006006006006e-05,
|
| 4064 |
+
"loss": 0.3117,
|
| 4065 |
+
"step": 6600
|
| 4066 |
+
},
|
| 4067 |
+
{
|
| 4068 |
+
"epoch": 8.92,
|
| 4069 |
+
"learning_rate": 5.930930930930931e-05,
|
| 4070 |
+
"loss": 0.3402,
|
| 4071 |
+
"step": 6610
|
| 4072 |
+
},
|
| 4073 |
+
{
|
| 4074 |
+
"epoch": 8.94,
|
| 4075 |
+
"learning_rate": 5.855855855855856e-05,
|
| 4076 |
+
"loss": 0.3251,
|
| 4077 |
+
"step": 6620
|
| 4078 |
+
},
|
| 4079 |
+
{
|
| 4080 |
+
"epoch": 8.95,
|
| 4081 |
+
"learning_rate": 5.780780780780781e-05,
|
| 4082 |
+
"loss": 0.3538,
|
| 4083 |
+
"step": 6630
|
| 4084 |
+
},
|
| 4085 |
+
{
|
| 4086 |
+
"epoch": 8.96,
|
| 4087 |
+
"learning_rate": 5.7057057057057056e-05,
|
| 4088 |
+
"loss": 0.3323,
|
| 4089 |
+
"step": 6640
|
| 4090 |
+
},
|
| 4091 |
+
{
|
| 4092 |
+
"epoch": 8.98,
|
| 4093 |
+
"learning_rate": 5.630630630630631e-05,
|
| 4094 |
+
"loss": 0.2546,
|
| 4095 |
+
"step": 6650
|
| 4096 |
+
},
|
| 4097 |
+
{
|
| 4098 |
+
"epoch": 8.99,
|
| 4099 |
+
"learning_rate": 5.555555555555555e-05,
|
| 4100 |
+
"loss": 0.3567,
|
| 4101 |
+
"step": 6660
|
| 4102 |
+
},
|
| 4103 |
+
{
|
| 4104 |
+
"epoch": 9.0,
|
| 4105 |
+
"eval_accuracy": 0.9256851370274055,
|
| 4106 |
+
"eval_f1": 0.9250707611820068,
|
| 4107 |
+
"eval_loss": 0.27885687351226807,
|
| 4108 |
+
"eval_precision": 0.9262127050820336,
|
| 4109 |
+
"eval_recall": 0.9250555525343819,
|
| 4110 |
+
"eval_runtime": 194.8139,
|
| 4111 |
+
"eval_samples_per_second": 51.321,
|
| 4112 |
+
"eval_steps_per_second": 1.904,
|
| 4113 |
+
"step": 6666
|
| 4114 |
+
},
|
| 4115 |
+
{
|
| 4116 |
+
"epoch": 9.0,
|
| 4117 |
+
"learning_rate": 5.480480480480481e-05,
|
| 4118 |
+
"loss": 0.2732,
|
| 4119 |
+
"step": 6670
|
| 4120 |
+
},
|
| 4121 |
+
{
|
| 4122 |
+
"epoch": 9.02,
|
| 4123 |
+
"learning_rate": 5.405405405405406e-05,
|
| 4124 |
+
"loss": 0.316,
|
| 4125 |
+
"step": 6680
|
| 4126 |
+
},
|
| 4127 |
+
{
|
| 4128 |
+
"epoch": 9.03,
|
| 4129 |
+
"learning_rate": 5.3303303303303305e-05,
|
| 4130 |
+
"loss": 0.2812,
|
| 4131 |
+
"step": 6690
|
| 4132 |
+
},
|
| 4133 |
+
{
|
| 4134 |
+
"epoch": 9.04,
|
| 4135 |
+
"learning_rate": 5.2552552552552556e-05,
|
| 4136 |
+
"loss": 0.3061,
|
| 4137 |
+
"step": 6700
|
| 4138 |
+
},
|
| 4139 |
+
{
|
| 4140 |
+
"epoch": 9.06,
|
| 4141 |
+
"learning_rate": 5.18018018018018e-05,
|
| 4142 |
+
"loss": 0.2993,
|
| 4143 |
+
"step": 6710
|
| 4144 |
+
},
|
| 4145 |
+
{
|
| 4146 |
+
"epoch": 9.07,
|
| 4147 |
+
"learning_rate": 5.105105105105105e-05,
|
| 4148 |
+
"loss": 0.323,
|
| 4149 |
+
"step": 6720
|
| 4150 |
+
},
|
| 4151 |
+
{
|
| 4152 |
+
"epoch": 9.09,
|
| 4153 |
+
"learning_rate": 5.03003003003003e-05,
|
| 4154 |
+
"loss": 0.2922,
|
| 4155 |
+
"step": 6730
|
| 4156 |
+
},
|
| 4157 |
+
{
|
| 4158 |
+
"epoch": 9.1,
|
| 4159 |
+
"learning_rate": 4.954954954954955e-05,
|
| 4160 |
+
"loss": 0.3282,
|
| 4161 |
+
"step": 6740
|
| 4162 |
+
},
|
| 4163 |
+
{
|
| 4164 |
+
"epoch": 9.11,
|
| 4165 |
+
"learning_rate": 4.87987987987988e-05,
|
| 4166 |
+
"loss": 0.2847,
|
| 4167 |
+
"step": 6750
|
| 4168 |
+
},
|
| 4169 |
+
{
|
| 4170 |
+
"epoch": 9.13,
|
| 4171 |
+
"learning_rate": 4.804804804804805e-05,
|
| 4172 |
+
"loss": 0.2945,
|
| 4173 |
+
"step": 6760
|
| 4174 |
+
},
|
| 4175 |
+
{
|
| 4176 |
+
"epoch": 9.14,
|
| 4177 |
+
"learning_rate": 4.72972972972973e-05,
|
| 4178 |
+
"loss": 0.2525,
|
| 4179 |
+
"step": 6770
|
| 4180 |
+
},
|
| 4181 |
+
{
|
| 4182 |
+
"epoch": 9.15,
|
| 4183 |
+
"learning_rate": 4.6546546546546544e-05,
|
| 4184 |
+
"loss": 0.2907,
|
| 4185 |
+
"step": 6780
|
| 4186 |
+
},
|
| 4187 |
+
{
|
| 4188 |
+
"epoch": 9.17,
|
| 4189 |
+
"learning_rate": 4.57957957957958e-05,
|
| 4190 |
+
"loss": 0.3036,
|
| 4191 |
+
"step": 6790
|
| 4192 |
+
},
|
| 4193 |
+
{
|
| 4194 |
+
"epoch": 9.18,
|
| 4195 |
+
"learning_rate": 4.5045045045045046e-05,
|
| 4196 |
+
"loss": 0.2495,
|
| 4197 |
+
"step": 6800
|
| 4198 |
+
},
|
| 4199 |
+
{
|
| 4200 |
+
"epoch": 9.19,
|
| 4201 |
+
"learning_rate": 4.42942942942943e-05,
|
| 4202 |
+
"loss": 0.2752,
|
| 4203 |
+
"step": 6810
|
| 4204 |
+
},
|
| 4205 |
+
{
|
| 4206 |
+
"epoch": 9.21,
|
| 4207 |
+
"learning_rate": 4.354354354354354e-05,
|
| 4208 |
+
"loss": 0.3388,
|
| 4209 |
+
"step": 6820
|
| 4210 |
+
},
|
| 4211 |
+
{
|
| 4212 |
+
"epoch": 9.22,
|
| 4213 |
+
"learning_rate": 4.279279279279279e-05,
|
| 4214 |
+
"loss": 0.3166,
|
| 4215 |
+
"step": 6830
|
| 4216 |
+
},
|
| 4217 |
+
{
|
| 4218 |
+
"epoch": 9.23,
|
| 4219 |
+
"learning_rate": 4.204204204204204e-05,
|
| 4220 |
+
"loss": 0.3494,
|
| 4221 |
+
"step": 6840
|
| 4222 |
+
},
|
| 4223 |
+
{
|
| 4224 |
+
"epoch": 9.25,
|
| 4225 |
+
"learning_rate": 4.1291291291291294e-05,
|
| 4226 |
+
"loss": 0.2602,
|
| 4227 |
+
"step": 6850
|
| 4228 |
+
},
|
| 4229 |
+
{
|
| 4230 |
+
"epoch": 9.26,
|
| 4231 |
+
"learning_rate": 4.0540540540540545e-05,
|
| 4232 |
+
"loss": 0.2571,
|
| 4233 |
+
"step": 6860
|
| 4234 |
+
},
|
| 4235 |
+
{
|
| 4236 |
+
"epoch": 9.27,
|
| 4237 |
+
"learning_rate": 3.978978978978979e-05,
|
| 4238 |
+
"loss": 0.3087,
|
| 4239 |
+
"step": 6870
|
| 4240 |
+
},
|
| 4241 |
+
{
|
| 4242 |
+
"epoch": 9.29,
|
| 4243 |
+
"learning_rate": 3.903903903903904e-05,
|
| 4244 |
+
"loss": 0.2927,
|
| 4245 |
+
"step": 6880
|
| 4246 |
+
},
|
| 4247 |
+
{
|
| 4248 |
+
"epoch": 9.3,
|
| 4249 |
+
"learning_rate": 3.8288288288288285e-05,
|
| 4250 |
+
"loss": 0.2868,
|
| 4251 |
+
"step": 6890
|
| 4252 |
+
},
|
| 4253 |
+
{
|
| 4254 |
+
"epoch": 9.31,
|
| 4255 |
+
"learning_rate": 3.7537537537537536e-05,
|
| 4256 |
+
"loss": 0.2722,
|
| 4257 |
+
"step": 6900
|
| 4258 |
+
},
|
| 4259 |
+
{
|
| 4260 |
+
"epoch": 9.33,
|
| 4261 |
+
"learning_rate": 3.678678678678679e-05,
|
| 4262 |
+
"loss": 0.249,
|
| 4263 |
+
"step": 6910
|
| 4264 |
+
},
|
| 4265 |
+
{
|
| 4266 |
+
"epoch": 9.34,
|
| 4267 |
+
"learning_rate": 3.603603603603604e-05,
|
| 4268 |
+
"loss": 0.2773,
|
| 4269 |
+
"step": 6920
|
| 4270 |
+
},
|
| 4271 |
+
{
|
| 4272 |
+
"epoch": 9.36,
|
| 4273 |
+
"learning_rate": 3.528528528528529e-05,
|
| 4274 |
+
"loss": 0.2947,
|
| 4275 |
+
"step": 6930
|
| 4276 |
+
},
|
| 4277 |
+
{
|
| 4278 |
+
"epoch": 9.37,
|
| 4279 |
+
"learning_rate": 3.453453453453453e-05,
|
| 4280 |
+
"loss": 0.3421,
|
| 4281 |
+
"step": 6940
|
| 4282 |
+
},
|
| 4283 |
+
{
|
| 4284 |
+
"epoch": 9.38,
|
| 4285 |
+
"learning_rate": 3.3783783783783784e-05,
|
| 4286 |
+
"loss": 0.2547,
|
| 4287 |
+
"step": 6950
|
| 4288 |
+
},
|
| 4289 |
+
{
|
| 4290 |
+
"epoch": 9.4,
|
| 4291 |
+
"learning_rate": 3.303303303303303e-05,
|
| 4292 |
+
"loss": 0.2803,
|
| 4293 |
+
"step": 6960
|
| 4294 |
+
},
|
| 4295 |
+
{
|
| 4296 |
+
"epoch": 9.41,
|
| 4297 |
+
"learning_rate": 3.2282282282282286e-05,
|
| 4298 |
+
"loss": 0.2675,
|
| 4299 |
+
"step": 6970
|
| 4300 |
+
},
|
| 4301 |
+
{
|
| 4302 |
+
"epoch": 9.42,
|
| 4303 |
+
"learning_rate": 3.153153153153153e-05,
|
| 4304 |
+
"loss": 0.2654,
|
| 4305 |
+
"step": 6980
|
| 4306 |
+
},
|
| 4307 |
+
{
|
| 4308 |
+
"epoch": 9.44,
|
| 4309 |
+
"learning_rate": 3.078078078078078e-05,
|
| 4310 |
+
"loss": 0.2709,
|
| 4311 |
+
"step": 6990
|
| 4312 |
+
},
|
| 4313 |
+
{
|
| 4314 |
+
"epoch": 9.45,
|
| 4315 |
+
"learning_rate": 3.003003003003003e-05,
|
| 4316 |
+
"loss": 0.2359,
|
| 4317 |
+
"step": 7000
|
| 4318 |
+
},
|
| 4319 |
+
{
|
| 4320 |
+
"epoch": 9.46,
|
| 4321 |
+
"learning_rate": 2.927927927927928e-05,
|
| 4322 |
+
"loss": 0.2535,
|
| 4323 |
+
"step": 7010
|
| 4324 |
+
},
|
| 4325 |
+
{
|
| 4326 |
+
"epoch": 9.48,
|
| 4327 |
+
"learning_rate": 2.8528528528528528e-05,
|
| 4328 |
+
"loss": 0.2867,
|
| 4329 |
+
"step": 7020
|
| 4330 |
+
},
|
| 4331 |
+
{
|
| 4332 |
+
"epoch": 9.49,
|
| 4333 |
+
"learning_rate": 2.7777777777777776e-05,
|
| 4334 |
+
"loss": 0.254,
|
| 4335 |
+
"step": 7030
|
| 4336 |
+
},
|
| 4337 |
+
{
|
| 4338 |
+
"epoch": 9.5,
|
| 4339 |
+
"learning_rate": 2.702702702702703e-05,
|
| 4340 |
+
"loss": 0.2382,
|
| 4341 |
+
"step": 7040
|
| 4342 |
+
},
|
| 4343 |
+
{
|
| 4344 |
+
"epoch": 9.52,
|
| 4345 |
+
"learning_rate": 2.6276276276276278e-05,
|
| 4346 |
+
"loss": 0.2711,
|
| 4347 |
+
"step": 7050
|
| 4348 |
+
},
|
| 4349 |
+
{
|
| 4350 |
+
"epoch": 9.53,
|
| 4351 |
+
"learning_rate": 2.5525525525525525e-05,
|
| 4352 |
+
"loss": 0.2785,
|
| 4353 |
+
"step": 7060
|
| 4354 |
+
},
|
| 4355 |
+
{
|
| 4356 |
+
"epoch": 9.54,
|
| 4357 |
+
"learning_rate": 2.4774774774774777e-05,
|
| 4358 |
+
"loss": 0.2695,
|
| 4359 |
+
"step": 7070
|
| 4360 |
+
},
|
| 4361 |
+
{
|
| 4362 |
+
"epoch": 9.56,
|
| 4363 |
+
"learning_rate": 2.4024024024024024e-05,
|
| 4364 |
+
"loss": 0.3143,
|
| 4365 |
+
"step": 7080
|
| 4366 |
+
},
|
| 4367 |
+
{
|
| 4368 |
+
"epoch": 9.57,
|
| 4369 |
+
"learning_rate": 2.3273273273273272e-05,
|
| 4370 |
+
"loss": 0.2781,
|
| 4371 |
+
"step": 7090
|
| 4372 |
+
},
|
| 4373 |
+
{
|
| 4374 |
+
"epoch": 9.58,
|
| 4375 |
+
"learning_rate": 2.2522522522522523e-05,
|
| 4376 |
+
"loss": 0.254,
|
| 4377 |
+
"step": 7100
|
| 4378 |
+
},
|
| 4379 |
+
{
|
| 4380 |
+
"epoch": 9.6,
|
| 4381 |
+
"learning_rate": 2.177177177177177e-05,
|
| 4382 |
+
"loss": 0.2371,
|
| 4383 |
+
"step": 7110
|
| 4384 |
+
},
|
| 4385 |
+
{
|
| 4386 |
+
"epoch": 9.61,
|
| 4387 |
+
"learning_rate": 2.102102102102102e-05,
|
| 4388 |
+
"loss": 0.2558,
|
| 4389 |
+
"step": 7120
|
| 4390 |
+
},
|
| 4391 |
+
{
|
| 4392 |
+
"epoch": 9.63,
|
| 4393 |
+
"learning_rate": 2.0270270270270273e-05,
|
| 4394 |
+
"loss": 0.2748,
|
| 4395 |
+
"step": 7130
|
| 4396 |
+
},
|
| 4397 |
+
{
|
| 4398 |
+
"epoch": 9.64,
|
| 4399 |
+
"learning_rate": 1.951951951951952e-05,
|
| 4400 |
+
"loss": 0.3006,
|
| 4401 |
+
"step": 7140
|
| 4402 |
+
},
|
| 4403 |
+
{
|
| 4404 |
+
"epoch": 9.65,
|
| 4405 |
+
"learning_rate": 1.8768768768768768e-05,
|
| 4406 |
+
"loss": 0.3219,
|
| 4407 |
+
"step": 7150
|
| 4408 |
+
},
|
| 4409 |
+
{
|
| 4410 |
+
"epoch": 9.67,
|
| 4411 |
+
"learning_rate": 1.801801801801802e-05,
|
| 4412 |
+
"loss": 0.3013,
|
| 4413 |
+
"step": 7160
|
| 4414 |
+
},
|
| 4415 |
+
{
|
| 4416 |
+
"epoch": 9.68,
|
| 4417 |
+
"learning_rate": 1.7267267267267267e-05,
|
| 4418 |
+
"loss": 0.2612,
|
| 4419 |
+
"step": 7170
|
| 4420 |
+
},
|
| 4421 |
+
{
|
| 4422 |
+
"epoch": 9.69,
|
| 4423 |
+
"learning_rate": 1.6516516516516514e-05,
|
| 4424 |
+
"loss": 0.2406,
|
| 4425 |
+
"step": 7180
|
| 4426 |
+
},
|
| 4427 |
+
{
|
| 4428 |
+
"epoch": 9.71,
|
| 4429 |
+
"learning_rate": 1.5765765765765765e-05,
|
| 4430 |
+
"loss": 0.2628,
|
| 4431 |
+
"step": 7190
|
| 4432 |
+
},
|
| 4433 |
+
{
|
| 4434 |
+
"epoch": 9.72,
|
| 4435 |
+
"learning_rate": 1.5015015015015015e-05,
|
| 4436 |
+
"loss": 0.2798,
|
| 4437 |
+
"step": 7200
|
| 4438 |
+
},
|
| 4439 |
+
{
|
| 4440 |
+
"epoch": 9.73,
|
| 4441 |
+
"learning_rate": 1.4264264264264264e-05,
|
| 4442 |
+
"loss": 0.2725,
|
| 4443 |
+
"step": 7210
|
| 4444 |
+
},
|
| 4445 |
+
{
|
| 4446 |
+
"epoch": 9.75,
|
| 4447 |
+
"learning_rate": 1.3513513513513515e-05,
|
| 4448 |
+
"loss": 0.3193,
|
| 4449 |
+
"step": 7220
|
| 4450 |
+
},
|
| 4451 |
+
{
|
| 4452 |
+
"epoch": 9.76,
|
| 4453 |
+
"learning_rate": 1.2762762762762763e-05,
|
| 4454 |
+
"loss": 0.277,
|
| 4455 |
+
"step": 7230
|
| 4456 |
+
},
|
| 4457 |
+
{
|
| 4458 |
+
"epoch": 9.77,
|
| 4459 |
+
"learning_rate": 1.2012012012012012e-05,
|
| 4460 |
+
"loss": 0.3349,
|
| 4461 |
+
"step": 7240
|
| 4462 |
+
},
|
| 4463 |
+
{
|
| 4464 |
+
"epoch": 9.79,
|
| 4465 |
+
"learning_rate": 1.1261261261261261e-05,
|
| 4466 |
+
"loss": 0.272,
|
| 4467 |
+
"step": 7250
|
| 4468 |
+
},
|
| 4469 |
+
{
|
| 4470 |
+
"epoch": 9.8,
|
| 4471 |
+
"learning_rate": 1.051051051051051e-05,
|
| 4472 |
+
"loss": 0.3017,
|
| 4473 |
+
"step": 7260
|
| 4474 |
+
},
|
| 4475 |
+
{
|
| 4476 |
+
"epoch": 9.81,
|
| 4477 |
+
"learning_rate": 9.75975975975976e-06,
|
| 4478 |
+
"loss": 0.3224,
|
| 4479 |
+
"step": 7270
|
| 4480 |
+
},
|
| 4481 |
+
{
|
| 4482 |
+
"epoch": 9.83,
|
| 4483 |
+
"learning_rate": 9.00900900900901e-06,
|
| 4484 |
+
"loss": 0.2935,
|
| 4485 |
+
"step": 7280
|
| 4486 |
+
},
|
| 4487 |
+
{
|
| 4488 |
+
"epoch": 9.84,
|
| 4489 |
+
"learning_rate": 8.258258258258257e-06,
|
| 4490 |
+
"loss": 0.3441,
|
| 4491 |
+
"step": 7290
|
| 4492 |
+
},
|
| 4493 |
+
{
|
| 4494 |
+
"epoch": 9.85,
|
| 4495 |
+
"learning_rate": 7.507507507507507e-06,
|
| 4496 |
+
"loss": 0.3263,
|
| 4497 |
+
"step": 7300
|
| 4498 |
+
},
|
| 4499 |
+
{
|
| 4500 |
+
"epoch": 9.87,
|
| 4501 |
+
"learning_rate": 6.7567567567567575e-06,
|
| 4502 |
+
"loss": 0.2503,
|
| 4503 |
+
"step": 7310
|
| 4504 |
+
},
|
| 4505 |
+
{
|
| 4506 |
+
"epoch": 9.88,
|
| 4507 |
+
"learning_rate": 6.006006006006006e-06,
|
| 4508 |
+
"loss": 0.316,
|
| 4509 |
+
"step": 7320
|
| 4510 |
+
},
|
| 4511 |
+
{
|
| 4512 |
+
"epoch": 9.9,
|
| 4513 |
+
"learning_rate": 5.255255255255255e-06,
|
| 4514 |
+
"loss": 0.309,
|
| 4515 |
+
"step": 7330
|
| 4516 |
+
},
|
| 4517 |
+
{
|
| 4518 |
+
"epoch": 9.91,
|
| 4519 |
+
"learning_rate": 4.504504504504505e-06,
|
| 4520 |
+
"loss": 0.2497,
|
| 4521 |
+
"step": 7340
|
| 4522 |
+
},
|
| 4523 |
+
{
|
| 4524 |
+
"epoch": 9.92,
|
| 4525 |
+
"learning_rate": 3.7537537537537537e-06,
|
| 4526 |
+
"loss": 0.206,
|
| 4527 |
+
"step": 7350
|
| 4528 |
+
},
|
| 4529 |
+
{
|
| 4530 |
+
"epoch": 9.94,
|
| 4531 |
+
"learning_rate": 3.003003003003003e-06,
|
| 4532 |
+
"loss": 0.2813,
|
| 4533 |
+
"step": 7360
|
| 4534 |
+
},
|
| 4535 |
+
{
|
| 4536 |
+
"epoch": 9.95,
|
| 4537 |
+
"learning_rate": 2.2522522522522524e-06,
|
| 4538 |
+
"loss": 0.3494,
|
| 4539 |
+
"step": 7370
|
| 4540 |
+
},
|
| 4541 |
+
{
|
| 4542 |
+
"epoch": 9.96,
|
| 4543 |
+
"learning_rate": 1.5015015015015015e-06,
|
| 4544 |
+
"loss": 0.3268,
|
| 4545 |
+
"step": 7380
|
| 4546 |
+
},
|
| 4547 |
+
{
|
| 4548 |
+
"epoch": 9.98,
|
| 4549 |
+
"learning_rate": 7.507507507507508e-07,
|
| 4550 |
+
"loss": 0.2431,
|
| 4551 |
+
"step": 7390
|
| 4552 |
+
},
|
| 4553 |
+
{
|
| 4554 |
+
"epoch": 9.99,
|
| 4555 |
+
"learning_rate": 0.0,
|
| 4556 |
+
"loss": 0.3313,
|
| 4557 |
+
"step": 7400
|
| 4558 |
+
},
|
| 4559 |
+
{
|
| 4560 |
+
"epoch": 9.99,
|
| 4561 |
+
"eval_accuracy": 0.9310862172434486,
|
| 4562 |
+
"eval_f1": 0.9305170890009876,
|
| 4563 |
+
"eval_loss": 0.26112696528434753,
|
| 4564 |
+
"eval_precision": 0.9311947598825216,
|
| 4565 |
+
"eval_recall": 0.9304846180354004,
|
| 4566 |
+
"eval_runtime": 191.2068,
|
| 4567 |
+
"eval_samples_per_second": 52.289,
|
| 4568 |
+
"eval_steps_per_second": 1.94,
|
| 4569 |
+
"step": 7400
|
| 4570 |
+
},
|
| 4571 |
+
{
|
| 4572 |
+
"epoch": 9.99,
|
| 4573 |
+
"step": 7400,
|
| 4574 |
+
"total_flos": 1.8619741375601993e+20,
|
| 4575 |
+
"train_loss": 0.7571127768142804,
|
| 4576 |
+
"train_runtime": 37998.3887,
|
| 4577 |
+
"train_samples_per_second": 21.049,
|
| 4578 |
+
"train_steps_per_second": 0.195
|
| 4579 |
+
}
|
| 4580 |
+
],
|
| 4581 |
+
"logging_steps": 10,
|
| 4582 |
+
"max_steps": 7400,
|
| 4583 |
+
"num_train_epochs": 10,
|
| 4584 |
+
"save_steps": 500,
|
| 4585 |
+
"total_flos": 1.8619741375601993e+20,
|
| 4586 |
+
"trial_name": null,
|
| 4587 |
+
"trial_params": null
|
| 4588 |
+
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 4155
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38a2a1afe088d2ccfd529d760beec1122a715087474e1e90415ae300eef2d1d5
|
| 3 |
size 4155
|