Enable flash_attention_2 support since the underlying Mistral model supports it
#3
by
winglian
- opened
No description provided.
lievan
changed pull request status to
merged