Add training logs and README
Browse files
README.md
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Fine-tuned Model: Judge-Adaptor-v3
|
| 2 |
+
|
| 3 |
+
## 📚 Training Configuration
|
| 4 |
+
- **data_path**: `QomSSLab/Legal_SyntheticDraftRuling_Selected_v2`
|
| 5 |
+
- **output_dir**: `gemma312b_lora_chckpnts`
|
| 6 |
+
- **new_model_name**: `Judge-Adaptor-v3`
|
| 7 |
+
- **data_ratio**: `1.0`
|
| 8 |
+
- **model_name**: `QomSSLab/Legal-gemma3-12b-it-lora-thinking`
|
| 9 |
+
- **use_4bit**: `False`
|
| 10 |
+
- **use_lora**: `True`
|
| 11 |
+
- **max_seq_length**: `40000`
|
| 12 |
+
- **batch_size**: `1`
|
| 13 |
+
- **gradient_accu**: `8`
|
| 14 |
+
- **epochs**: `2`
|
| 15 |
+
- **learning_rate**: `1e-05`
|
| 16 |
+
- **lora_alpha**: `64`
|
| 17 |
+
- **lora_drop**: `0.05`
|
| 18 |
+
- **lora_r**: `64`
|
| 19 |
+
- **tune_embedding_layer**: `False`
|
| 20 |
+
- **hf_token**: `********`
|
| 21 |
+
- **resume_from_checkpoint**: `False`
|
| 22 |
+
- **use_8bit_optimizer**: `True`
|
| 23 |
+
- **push_to_hub**: `True`
|
| 24 |
+
- **push_lora_only**: `True`
|
| 25 |
+
- **train_only_on_assistant**: `True`
|
| 26 |
+
- **last_response**: `False`
|
| 27 |
+
|
| 28 |
+
---
|
| 29 |
+
Auto-generated after training.
|