Update README.md
Browse files
README.md
CHANGED
|
@@ -8,7 +8,7 @@ tags:
|
|
| 8 |
|
| 9 |
This Mistral 7B fine-tune is trained (for 150 steps) to talk like a **human**, not a "helpful assistant"!
|
| 10 |
|
| 11 |
-
|
| 12 |
|
| 13 |
**Example with temperature = 0.8, repeat penality = 1.0, top k sampling = 40:**
|
| 14 |
|
|
|
|
| 8 |
|
| 9 |
This Mistral 7B fine-tune is trained (for 150 steps) to talk like a **human**, not a "helpful assistant"!
|
| 10 |
|
| 11 |
+
It's also *very* beta right now. The dataset ([qingy2024/Natural-Text-ShareGPT](https://huggingface.co/datasets/qingy2024/Natural-Text-ShareGPT)) can definitely be improved.
|
| 12 |
|
| 13 |
**Example with temperature = 0.8, repeat penality = 1.0, top k sampling = 40:**
|
| 14 |
|