Commit
•
88b9dbc
1
Parent(s):
876c817
Update README.md
Browse files
README.md
CHANGED
|
@@ -103,6 +103,7 @@ export use_triton=False
|
|
| 103 |
>>> pred = model.generate(**inputs, max_new_tokens=2048, repetition_penalty=1.0)
|
| 104 |
>>> print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
|
| 105 |
```
|
|
|
|
| 106 |
|
| 107 |
|
| 108 |
# Fine-tuning the Model
|
|
|
|
| 103 |
>>> pred = model.generate(**inputs, max_new_tokens=2048, repetition_penalty=1.0)
|
| 104 |
>>> print(tokenizer.decode(pred.cpu()[0], skip_special_tokens=True))
|
| 105 |
```
|
| 106 |
+
* **Note**: we recommend to use `bfloat16` in `TransNormerLLM`, `float16` might lead `nan` error, please check your divce compatibility!
|
| 107 |
|
| 108 |
|
| 109 |
# Fine-tuning the Model
|