Update AFMoE architecture to use v5-style MoE impl #44063
v5-style AFMoE impl
c3be2d5b
Merge branch 'main' of https://github.com/AutumnAurelium/transformers
20896f86
Merge branch 'huggingface:main' into main
a9f620a5
don't unnecessarily return router logits
359503b8
Merge branch 'main' of https://github.com/AutumnAurelium/transformers
690e9d38
winglian
approved these changes
on 2026-03-03
inherit MoE code and refactor for stylistic consistency
9467bd35
remove pointless type alias
6815d21e
Merge branch 'main' of https://github.com/huggingface/transformers
8f59135a
Merge branch 'main' into main
6372b303
remove legacy cache reference
bde5fa65
type and lint
622f2daf
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub