Update README.md
Browse files
README.md
CHANGED
|
@@ -21,8 +21,15 @@ Llama-3.1-Centaur-70B is a foundation model of cognition model that can predict
|
|
| 21 |
This is the low-rank adapter which runs with unsloth on a single 80GB GPU.
|
| 22 |
|
| 23 |
```python
|
| 24 |
-
import
|
| 25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
```
|
| 27 |
|
| 28 |
Alternatively, you can also directly use the [merged model](https://huggingface.co/marcelbinz/Llama-3.1-Centaur-70B).
|
|
|
|
| 21 |
This is the low-rank adapter which runs with unsloth on a single 80GB GPU.
|
| 22 |
|
| 23 |
```python
|
| 24 |
+
from unsloth import FastLanguageModel
|
| 25 |
+
|
| 26 |
+
model_name = "marcelbinz/Llama-3.1-Centaur-70B-adapter"
|
| 27 |
+
model, tokenizer = FastLanguageModel.from_pretrained(
|
| 28 |
+
model_name = args.model,
|
| 29 |
+
max_seq_length = 32768,
|
| 30 |
+
dtype = None,
|
| 31 |
+
load_in_4bit = True,
|
| 32 |
+
)
|
| 33 |
```
|
| 34 |
|
| 35 |
Alternatively, you can also directly use the [merged model](https://huggingface.co/marcelbinz/Llama-3.1-Centaur-70B).
|