DeepSpeed
Log the skipped optimizer step on gradient overflow
#8417
Merged

Log the skipped optimizer step on gradient overflow #8417

sfc-gh-abkulkarni
sfc-gh-abkulkarni Log the skipped optimizer step on gradient overflow
a99600f0
sfc-gh-abkulkarni sfc-gh-abkulkarni requested a review from tjruwase tjruwase 45 days ago
sfc-gh-abkulkarni sfc-gh-abkulkarni requested a review from tohtana tohtana 45 days ago
chatgpt-codex-connector
chatgpt-codex-connector commented on 2026-09-04
sfc-gh-abkulkarni Emit the overflow warning from rank zero only
546dbb0c
sfc-gh-abkulkarni Trim the overflow warning comments to the non-obvious gate
2ce0b9ab
sfc-gh-abkulkarni Report the skipped step from the static loss scaler
290275e4
sfc-gh-abkulkarni sfc-gh-abkulkarni requested a review from loadams loadams 45 days ago
sfc-gh-abkulkarni Drop the comment from update_scale
b971c0b8
pengdurice
pengdurice approved these changes on 2026-09-14
pengdurice pengdurice enabled auto-merge 35 days ago
sfc-gh-truwase Merge branch 'master' into fix/log-skipped-step-on-overflow
b184af50
sfc-gh-truwase sfc-gh-truwase enabled auto-merge 9 days ago
sfc-gh-truwase sfc-gh-truwase merged 500997a4 into master 9 days ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone