DeepSpeed
Fix ZeRO-3: Use per-param dtype for output buffers in _allgather_params_coalesced
#8073
Merged

Fix ZeRO-3: Use per-param dtype for output buffers in _allgather_params_coalesced #8073

albertvillanova
albertvillanova albertvillanova requested a review from tjruwase tjruwase 53 days ago
albertvillanova albertvillanova requested a review from tohtana tohtana 53 days ago
albertvillanova Use per-param dtype in _allgather_params_coalesced output buffers
c43a5a53
albertvillanova albertvillanova force pushed from b6007b5f to c43a5a53 53 days ago
albertvillanova albertvillanova changed the title Fix zero3: Use per-param dtype for output buffers in _allgather_params_coalesced Fix ZeRO-3: Use per-param dtype for output buffers in _allgather_params_coalesced 53 days ago
albertvillanova Merge branch 'master' into fix-8072
db88220a
tohtana
tohtana approved these changes on 2026-06-23
tohtana tohtana enabled auto-merge 47 days ago
tohtana tohtana merged b5b3fded into master 47 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone