Difference between model card with "code" and without?
#1
by
Yhyu13
- opened
HI,
Would you like to elaborate more on the difference between model with code in name and those without?
It seems code version achieve better math scores. https://github.com/microsoft/ToRA
But there is no 70b-code version out there yet?
Hi there,
ToRA-Code series are fine-tuned from CodeLLaMA (7B, 13B, 34B), while ToRA (no Code in name) series are fine-tuned from LLaMA-2 (7B, 13B, 70B).
zubingou
changed discussion status to
closed