transformers
[`distributed`] Divide the expert gradients once when `efsdp_size == 1`
#50292
Merged

[`distributed`] Divide the expert gradients once when `efsdp_size == 1` #50292

3outeille merged 5 commits into main from fix-ep-efsdp1-grad-division
3outeille
3outeille Divide the expert gradients once when efsdp_size == 1
78f529fc
HuggingFaceDocBuilderDev
3outeille Shorten the efsdp_size == 1 comment
d9f193b4
3outeille Move the efsdp_size == 1 comment under the if
a95f97b7
3outeille 3outeille requested a review from ArthurZucker ArthurZucker 2 days ago
3outeille Merge branch 'main' into fix-ep-efsdp1-grad-division
7f7edc0e
3outeille 3outeille marked this pull request as ready for review 2 days ago
3outeille Simplify the bf16 expert grads test docstrings
8b20e7a2
ArthurZucker
ArthurZucker approved these changes on 2026-11-05
3outeille 3outeille enabled auto-merge 2 days ago
3outeille 3outeille merged b3187c7f into main 2 days ago
3outeille 3outeille deleted the fix-ep-efsdp1-grad-division branch 2 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone