Commit
•
66db107
1
Parent(s):
a1b28cb
Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,23 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: en
|
| 3 |
+
license: mit
|
| 4 |
+
library_name: pytorch
|
| 5 |
+
---
|
| 6 |
+
# Knowledge Continuity Regularized Network
|
| 7 |
+
Trainer Hyperparameters:
|
| 8 |
+
- `lr` = 5e-05
|
| 9 |
+
- `per_device_batch_size` = 8
|
| 10 |
+
- `gradient_accumulation_steps` = 2
|
| 11 |
+
- `weight_decay` = 1e-09
|
| 12 |
+
- `seed` = 42
|
| 13 |
+
|
| 14 |
+
Regularization Hyperparameters
|
| 15 |
+
- `numerical stability denominator constant` = 0.001
|
| 16 |
+
- `lambda` = 0.01
|
| 17 |
+
- `alpha` = 2.0
|
| 18 |
+
- `beta` = 2.0
|
| 19 |
+
|
| 20 |
+
Extended Logs:
|
| 21 |
+
|
| 22 |
+
|eval_loss|eval_accuracy|epoch|
|
| 23 |
+
|--|--|--|
|