DeepSpeed
Skip sequence parallel operations during eval
#7821
Merged

Skip sequence parallel operations during eval #7821

tohtana merged 6 commits into deepspeedai:master from jp1924:sp_eval
jp1924
jp1924 Add early return for non-training mode in UlyssesSPAttentionHF
750a881c
jp1924 jp1924 requested a review from tjruwase tjruwase 209 days ago
jp1924 jp1924 requested a review from tohtana tohtana 209 days ago
jp1924
tohtana
jp1924
tohtana
jp1924
jp1924 tohtana: add option to disable SP in eval model
21700eac
jp1924 jp1924 requested a review from loadams loadams 208 days ago
jp1924 Merge branch 'master' into sp_eval
7b181c62
jp1924
jp1924
tohtana
jp1924
jp1924 Merge branch 'master' into sp_eval
fcbee475
tohtana
tohtana tohtana enabled auto-merge (squash) 206 days ago
tohtana fixed formatting
7734bd1d
tohtana Merge branch 'master' into sp_eval
abf86fab
tohtana
tohtana approved these changes on 2026-01-30
tohtana tohtana merged a972f7c6 into master 206 days ago
jp1924

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone