transformers
Keep AFMoE expert biases in FP32 on BF16 load
#50687
Merged

Keep AFMoE expert biases in FP32 on BF16 load #50687

annekethvij
annekethvij Preserve AFMoE normalization weights and expert biases on BF16 load
7fa1fdd3
github-actions
annekethvij Keep only AFMoE expert biases in FP32 on BF16 load
aac42e1a
annekethvij annekethvij changed the title Preserve AFMoE normalization weights and expert biases on BF16 load Keep AFMoE expert biases in FP32 on BF16 load 5 days ago
github-actions
Rocketknight1
annekethvij
annekethvij
Rocketknight1
Rocketknight1 approved these changes on 2026-11-17
Rocketknight1 Rocketknight1 enabled auto-merge 1 day ago
HuggingFaceDocBuilderDev
Rocketknight1 Rocketknight1 merged 2649c0d4 into main 1 day ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone