Update README.md
Browse files
README.md
CHANGED
|
@@ -57,6 +57,4 @@ This model is trained by OpenRLHF.
|
|
| 57 |
|
| 58 |
model.eval().to("cuda")
|
| 59 |
```
|
| 60 |
-
8.
|
| 61 |
-
- With FP32, the accuracy is:
|
| 62 |
-
- With BF16, the accuracy is:
|
|
|
|
| 57 |
|
| 58 |
model.eval().to("cuda")
|
| 59 |
```
|
| 60 |
+
8. Test accuracy: 0.764425, chosen reward: -0.032832, reject reward: -1.620852
|
|
|
|
|
|