Skip to content

Conversation

@rccchoudhury
Copy link

Following PR (); this re-enables using Flash Attention and bf16 precision for the internvideo2 encoder. I verified this does not affect downstream performance, and this increases the inference speed by ~4.7x.

@rccchoudhury rccchoudhury deleted the rcchoudhury/use_xformers_iv2 branch June 28, 2024 17:43
@rccchoudhury rccchoudhury restored the rcchoudhury/use_xformers_iv2 branch June 28, 2024 17:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants