Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,18 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
---
|
| 4 |
+
Lora finetuning on Wikipedia-10, applying counter factual data augmentation (CDA)
|
| 5 |
+
Dataset: Wikipedia-10
|
| 6 |
+
Target modules = ["q_proj", "k_proj", "v_proj", "dense", "fc1", "fc2"]
|
| 7 |
+
```
|
| 8 |
+
{
|
| 9 |
+
"epoch": 0.22326645805206993,
|
| 10 |
+
"total_flos": 1.04520375926784e+18,
|
| 11 |
+
"train_loss": 2.3526821104288103,
|
| 12 |
+
"train_runtime": 66080.1443,
|
| 13 |
+
"train_samples": 286653,
|
| 14 |
+
"train_samples_per_second": 0.969,
|
| 15 |
+
"train_steps_per_second": 0.03
|
| 16 |
+
}
|
| 17 |
+
```
|
| 18 |
+
Training script: https://github.com/ao9000/bias-bench/blob/main/experiments/run_clm.py
|