Spaces:
Running
on
Zero
Running
on
Zero
accelerate inference with dynamic AoT compilation and FA3
#12
by
linoyts
HF Staff
- opened
With
@cbensimon
and
@sayakpaul
we were able to accelerate inference with AoT compilation and FA3 (entirely loseless)
This PR makes the required changes to support this.
linoyts
changed pull request title from
accelerate inference + support dynamic compilation + fa3
to accelerate inference with dynamic AoT compilation and FA3
multimodalart
changed pull request status to
merged