Update README.md
Browse files
README.md
CHANGED
|
@@ -57,6 +57,17 @@ Tulu 2 13B is a fine-tuned version of Llama 2 that was trained on a mix of publi
|
|
| 57 |
| Claude 2 | - |RLHF |8.06| 91.36|
|
| 58 |
| GPT-4 | -| RLHF |8.99| 95.28|
|
| 59 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
|
| 61 |
## Intended uses & limitations
|
| 62 |
|
|
|
|
| 57 |
| Claude 2 | - |RLHF |8.06| 91.36|
|
| 58 |
| GPT-4 | -| RLHF |8.99| 95.28|
|
| 59 |
|
| 60 |
+
## Input Format
|
| 61 |
+
|
| 62 |
+
The model is trained to use the following format (note the newlines):
|
| 63 |
+
```
|
| 64 |
+
<|user|>
|
| 65 |
+
Your message here!
|
| 66 |
+
<|assistant|>
|
| 67 |
+
```
|
| 68 |
+
|
| 69 |
+
For best results, format all inputs in this manner. **Make sure to include a newline after `<|assistant|>`, this can affect generation quality quite a bit.**
|
| 70 |
+
|
| 71 |
|
| 72 |
## Intended uses & limitations
|
| 73 |
|