DeepSpeed
Mixed-precision: per-policy param/buffer dtype cast (preserve fp32 buffers)
#8066
Merged
Overview
Commits
8
Changes
View On GitHub

Mixed-precision: per-policy param/buffer dtype cast (preserve fp32 buffers) #8066

tjruwase merged 8 commits into master from mixed-precision-dtype
sfc-gh-truwase
sfc-gh-truwase Mixed-precision: per-policy param/buffer dtype cast (preserve fp32 bu…
9aee8a9b
sfc-gh-truwase sfc-gh-truwase requested a review from tjruwase tjruwase 7 days ago
sfc-gh-truwase sfc-gh-truwase requested a review from tohtana tohtana 7 days ago
chatgpt-codex-connector
chatgpt-codex-connector commented on 2026-06-15
sfc-gh-truwase Add tests for mixed-precision param/buffer dtype cast
9c34c91a
sfc-gh-truwase sfc-gh-truwase requested a review from loadams loadams 7 days ago
stas00
stas00 approved these changes on 2026-06-15
sfc-gh-truwase Reject param_dtype that conflicts with the enabled precision mode
5d4caad1
stas00
stas00 commented on 2026-06-15
tjruwase Update deepspeed/runtime/config.py
c2a7988b
tjruwase Merge branch 'master' into mixed-precision-dtype
9432ebc2
sfc-gh-truwase UT zero stages
8ba6b4ce
PKUWZP PKUWZP requested a review from PKUWZP PKUWZP 6 days ago
PKUWZP
stas00 format
b2f13071
stas00
stas00 commented on 2026-06-16
stas00 Apply suggestion from @stas00
86ab8988
PKUWZP
PKUWZP approved these changes on 2026-06-16
tjruwase tjruwase merged b919284a into master 6 days ago
tjruwase tjruwase deleted the mixed-precision-dtype branch 6 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum