DeepSpeed
AutoEP: one program per token in the fused weighted restore forward
#8678
Merged

AutoEP: one program per token in the fused weighted restore forward #8678

yh0903
yh0903 Give each fused weighted restore program a whole token
3c477efa
yh0903 Document whole-token restore numerics and guard against FP contraction
0ec88e95
yh0903 yh0903 marked this pull request as ready for review 20 days ago
yh0903 yh0903 requested a review from tohtana tohtana 20 days ago
yh0903 yh0903 requested a review from tjruwase tjruwase 20 days ago
yh0903 yh0903 requested a review from loadams loadams 20 days ago
hwchen2017 hwchen2017 requested a review from copilot-pull-request-reviewer copilot-pull-request-reviewer 18 days ago
copilot-pull-request-reviewer
copilot-pull-request-reviewer commented on 2026-09-29
tohtana
tohtana approved these changes on 2026-09-29
tohtana tohtana enabled auto-merge 18 days ago
pengdurice pengdurice added will-review
tohtana tohtana merged 11b24d21 into master 17 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone