Update README.md
Browse files
README.md
CHANGED
|
@@ -42,7 +42,6 @@ The fine-tuning process for Phi-3.5-mini-instruct-o1 employed the following tech
|
|
| 42 |
- **Compute Type:** BF16
|
| 43 |
- **Context Length:** 2048 tokens
|
| 44 |
- **Optimizer:** AdamW with cosine learning rate scheduling
|
| 45 |
-
- **Model Freezing:** Two trainable layers frozen
|
| 46 |
- **Additional Enhancement:** NEFTune with alpha 5
|
| 47 |
|
| 48 |
This fine-tuning approach was designed to efficiently adapt the model while maintaining its generalization capabilities and computational efficiency.
|
|
|
|
| 42 |
- **Compute Type:** BF16
|
| 43 |
- **Context Length:** 2048 tokens
|
| 44 |
- **Optimizer:** AdamW with cosine learning rate scheduling
|
|
|
|
| 45 |
- **Additional Enhancement:** NEFTune with alpha 5
|
| 46 |
|
| 47 |
This fine-tuning approach was designed to efficiently adapt the model while maintaining its generalization capabilities and computational efficiency.
|