Commit
•
3e53bc8
1
Parent(s):
b20c68b
Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Info
|
| 2 |
+
This is a V2 of the Gemini Nano V2 weights. The reason this is a V2 is the original conversion code was heavily bugged and extremelt slow. So Claude 3.5 Sonnet and o1-preview went in and fixed it! Now you'll notice the model has a lot more 2 dimension tensors and should, as a result, be easier to get working as a Gemma2 model!
|
| 3 |
+
|
| 4 |
+
# Known issues
|
| 5 |
+
The layer norms have an extra 2 dimensions. This will be fixed ASAP!
|