Skip to content

v0.0.4

Compare
Choose a tag to compare
@blefaudeux blefaudeux released this 17 Nov 04:53
· 939 commits to main since this release
1328ba7
  • Fixing causality not being respected by the scaled dot product attention
  • Fixing Favor causal trainability
  • Enabling FusedLayerNorm by default if Triton is available
  • Fixing Favor with fp16