DeepSpeed
Z1/2 init: flatten params on device
#7828
Merged

Z1/2 init: flatten params on device #7828

ksugama
ksugama ksugama force pushed from a07a21b7 to 293fbab4 112 days ago
ksugama ksugama changed the title Z1/2 Flatten Parameters on device Z1/2 init: flatten params on device 112 days ago
sfc-gh-truwase
tohtana Improve engine's cleanup (#7813)
500bd7f9
tohtana Ignore evoformer test (#7815)
792d52ca
nathon-lee Fix typos in accelerator setup guide (#7818)
374abb83
tohtana Raise clear error on in-place GatheredParameters edits without modifi…
c272de48
Flink-ddd [Bugfix] Resolve Rank index out of range during BWD when sp_size < wo…
d2aaf0f9
tohtana Update PyTorch to v2.9 for modal tests (#7816)
0bc2dd26
loadams Update version.txt to 0.18.6 after latest release (#7826)
4cb023c5
tohtana Fix leaf module race condition (#7825)
6064c2ae
jp1924 Skip sequence parallel operations during eval (#7821)
e86db608
tohtana Support custom partitioning patterns for AutoTP (#7806)
129b42c8
ksugama flatten gpu side
d307396d
ksugama repro script
6eb35e8e
ksugama detect gpu count in repro
48ecb1d7
ksugama add .venv to path
7f98cc85
ksugama clean up
b3944b4e
ksugama format and delete repro script
78e58fce
ksugama add dedicated test
7aa7073a
ksugama parametrize tests
85d670ad
sfc-gh-truwase Fix gradient is ready with z2 (#7829)
60d5cb97
tohtana Fix AutoTP custom patterns: respect use_default_specs (#7827)
36106318
ksugama ksugama force pushed from bf577e84 to 36106318 106 days ago
ksugama Merge branch 'master' into flatten-tensor-gpu
c2bb55b3
ksugama ksugama marked this pull request as ready for review 106 days ago
ksugama ksugama requested a review from tjruwase tjruwase 106 days ago
ksugama ksugama requested a review from tohtana tohtana 106 days ago
ksugama ksugama requested a review from loadams loadams 106 days ago
ksugama
ksugama commented on 2026-02-09
ksugama
ksugama Merge branch 'master' into flatten-tensor-gpu
489e1133
ksugama Merge branch 'master' into flatten-tensor-gpu
c3090a5a
ksugama Merge branch 'master' into flatten-tensor-gpu
496b3f77
ksugama
stas00
stas00
stas00
stas00 approved these changes on 2026-02-13
stas00 stas00 merged 84af8221 into master 102 days ago
ksugama
ksugama ksugama deleted the flatten-tensor-gpu branch 102 days ago
stas00

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone