Upload java KD Optuna tinyBERT checkpoint
Browse files
checkpoints/java/model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 57418068
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c16a90db410e6603230a8491fc0c74739072e668d267d845a793a00c480c350b
|
| 3 |
size 57418068
|
checkpoints/java/training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 5969
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cd31bce6baf17e054b8aaf2de5c39d7c07d533b729879c6ee07f4d295d8986b9
|
| 3 |
size 5969
|
checkpoints/java/trial_config.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
-
"lr":
|
| 3 |
"batch_size": 32,
|
| 4 |
-
"epochs":
|
| 5 |
-
"temperature": 3.
|
| 6 |
-
"lambda_hard": 0.
|
| 7 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"lr": 5.9146776044090886e-05,
|
| 3 |
"batch_size": 32,
|
| 4 |
+
"epochs": 19,
|
| 5 |
+
"temperature": 3.1492779350706006,
|
| 6 |
+
"lambda_hard": 0.3161226712833588
|
| 7 |
}
|
checkpoints/java/trial_metrics.json
CHANGED
|
@@ -1,27 +1,27 @@
|
|
| 1 |
{
|
| 2 |
-
"eval_loss":
|
| 3 |
-
"eval_macro_f1": 0.
|
| 4 |
-
"eval_micro_f1": 0.
|
| 5 |
-
"eval_macro_precision": 0.
|
| 6 |
-
"eval_macro_recall": 0.
|
| 7 |
"eval_thresholds": {
|
| 8 |
-
"summary": 0.
|
| 9 |
"Ownership": 0.4800000000000001,
|
| 10 |
-
"Expand": 0.
|
| 11 |
-
"usage": 0.
|
| 12 |
-
"Pointer": 0.
|
| 13 |
-
"deprecation": 0.
|
| 14 |
-
"rational": 0.
|
| 15 |
},
|
| 16 |
-
"eval_f1_summary": 0.
|
| 17 |
"eval_f1_Ownership": 1.0,
|
| 18 |
-
"eval_f1_Expand": 0.
|
| 19 |
-
"eval_f1_usage": 0.
|
| 20 |
-
"eval_f1_Pointer": 0.
|
| 21 |
"eval_f1_deprecation": 0.8235294117647058,
|
| 22 |
-
"eval_f1_rational": 0.
|
| 23 |
-
"eval_runtime": 0.
|
| 24 |
-
"eval_samples_per_second":
|
| 25 |
-
"eval_steps_per_second": 97.
|
| 26 |
-
"epoch":
|
| 27 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"eval_loss": 4.87105131149292,
|
| 3 |
+
"eval_macro_f1": 0.7406806320656656,
|
| 4 |
+
"eval_micro_f1": 0.800314465408805,
|
| 5 |
+
"eval_macro_precision": 0.7408403968173257,
|
| 6 |
+
"eval_macro_recall": 0.7679404509784634,
|
| 7 |
"eval_thresholds": {
|
| 8 |
+
"summary": 0.44000000000000006,
|
| 9 |
"Ownership": 0.4800000000000001,
|
| 10 |
+
"Expand": 0.44000000000000006,
|
| 11 |
+
"usage": 0.5800000000000001,
|
| 12 |
+
"Pointer": 0.5800000000000001,
|
| 13 |
+
"deprecation": 0.54,
|
| 14 |
+
"rational": 0.42000000000000004
|
| 15 |
},
|
| 16 |
+
"eval_f1_summary": 0.8741092636579573,
|
| 17 |
"eval_f1_Ownership": 1.0,
|
| 18 |
+
"eval_f1_Expand": 0.42105263157894735,
|
| 19 |
+
"eval_f1_usage": 0.8657243816254417,
|
| 20 |
+
"eval_f1_Pointer": 0.8648648648648649,
|
| 21 |
"eval_f1_deprecation": 0.8235294117647058,
|
| 22 |
+
"eval_f1_rational": 0.33548387096774196,
|
| 23 |
+
"eval_runtime": 0.3881,
|
| 24 |
+
"eval_samples_per_second": 3094.305,
|
| 25 |
+
"eval_steps_per_second": 97.905,
|
| 26 |
+
"epoch": 19.0
|
| 27 |
}
|