Open ternaus opened 11 months ago
It looks like Flash attention 2.0 brings solid improvements over 1.0.
It would be great to see an example of how one can leverage it in Stable Diffusion:
https://github.com/Dao-AILab/flash-attention
It looks like Flash attention 2.0 brings solid improvements over 1.0.
It would be great to see an example of how one can leverage it in Stable Diffusion:
https://github.com/Dao-AILab/flash-attention