Skip to content

Commit

Permalink
~7% speedup (1.57 to 1.69it/s) from switch to += in ldm.modules.atten…
Browse files Browse the repository at this point in the history 
…tion. (invoke-ai#482)

Tested on 8GB eGPU nvidia setup so YMMV.
512x512 output, max VRAM stays same.
  • Loading branch information
@mh-dm @lstein
mh-dm authored and lstein committed Sep 12, 2022
1 parent 32c5bf2 commit 68129ad
Showing 0 changed files with 0 additions and 0 deletions.

0 comments on commit 68129ad

Please sign in to comment.