DeepSpeed
Fix silent gradient-bucket drop on CPU in ZeRO-1/2 non-contiguous reduction
#8382
Merged

Fix silent gradient-bucket drop on CPU in ZeRO-1/2 non-contiguous reduction #8382

delock
delock delock requested a review from tjruwase tjruwase 43 days ago
delock delock requested a review from tohtana tohtana 43 days ago
delock delock requested a review from loadams loadams 43 days ago
chatgpt-codex-connector
chatgpt-codex-connector commented on 2026-09-01
delock Fix silent gradient-bucket drop on CPU and CPU-invalid device in test…
ad3c27e5
delock delock force pushed from 3757a3b9 to ad3c27e5 43 days ago
ebarkhordar
ebarkhordar commented on 2026-09-02
delock Exclude sparse gradient layouts from ZeRO-1/2 reduction buckets
bfddd3c9
delock Merge branch 'master' into fix/cpu-gradient-bucket-drop
03adfbb3
delock Merge branch 'master' into fix/cpu-gradient-bucket-drop
f3b9a5be
delock Merge branch 'master' into fix/cpu-gradient-bucket-drop
0bfd34e2
delock Exclude non-strided tensor layouts from ZeRO-1/2 reduction buckets
6040bbc6
delock Merge branch 'master' into fix/cpu-gradient-bucket-drop
b982351f
pengdurice
pengdurice approved these changes on 2026-09-18
delock delock merged 20a60f22 into master 23 days ago
delock delock deleted the fix/cpu-gradient-bucket-drop branch 23 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone