deepspeedai/DeepSpeed

Pull Requests Commits

Fix AutoTP custom patterns: respect use_default_specs (#7827)

tohtana committed 21 hours ago

Verified a44fb58f

Fix gradient is ready with z2 (#7829)

sfc-gh-truwase committed 4 days ago

Verified 0ccb2bb6

Support custom partitioning patterns for AutoTP (#7806)

tohtana committed 7 days ago

Verified 6b9cab1d

Skip sequence parallel operations during eval (#7821)

jp1924 committed 8 days ago

Verified a972f7c6

Fix leaf module race condition (#7825)

tohtana committed 8 days ago

Verified 52b1d4d6

Update version.txt to 0.18.6 after latest release (#7826)

loadams committed 8 days ago

Verified 7b15a6a7

Update PyTorch to v2.9 for modal tests (#7816)

tohtana committed 9 days ago

Verified b19987c7

[Bugfix] Resolve Rank index out of range during BWD when sp_size < world_size in Ulysses (#7809)

Flink-ddd committed 10 days ago

Verified bb250a25

Raise clear error on in-place GatheredParameters edits without modifier_rank (#7817)

tohtana committed 11 days ago

Verified d9f3d40a

Fix typos in accelerator setup guide (#7818)

nathon-lee committed 11 days ago

Verified 7b763459

Ignore evoformer test (#7815)

tohtana committed 13 days ago

Verified c89e0db8

Improve engine's cleanup (#7813)

tohtana committed 13 days ago

Verified 827767a4

Run Evoformer tests sequentially (#7810)

tohtana committed 14 days ago

Verified 156dcb2d

Fix ZeRO stage to choose BF16 optimizer in test (#7803)

tohtana committed 15 days ago

Verified 5b2ccad9

Fix ping-pong buffer index reset and removing redundant stream sync (#7805)

undersilence committed 16 days ago

Verified 15ad92b4

Fix Muon optimizer module path (#7802)

tohtana committed 17 days ago

Verified 3bc882fa

Add full test suite workflow (#7795)

tohtana committed 18 days ago

Verified 5aa2d17d

Fix BF16_Optimizer being used without ZeRO (#7790)

tohtana committed 18 days ago

Verified 09885efe

Fix issue with BF16 optimizer selection (#7788)

tohtana committed 18 days ago

Verified 1393f75d

Skip empty parameters in gradient reduction (#7789)

tohtana committed 18 days ago

Verified 114f971c

Fix backward for pipeline engine (#7787)

tohtana committed 18 days ago

Verified 3d896d7e

Fix AutoTP test numerical tolerance with rtol (#7794)

tohtana committed 19 days ago

Verified 374f6d09

Add sequential allgather optimization for ZeRO-3 (#7661)

aeeeeeep committed 19 days ago

Verified 60876090

fix checkpointing/loading of z0+bf16 (#7786)

tohtana committed 19 days ago

Verified 991ebd72

Fix Evoformer compilation (#7760)

sdvillal committed 20 days ago

Verified 43125a75

Fix Ulysses PEFT test (#7784)

tohtana committed 21 days ago

Verified 491a38ce

Fix gradient checkpointing with use_reentrant=True / PyTorch-style backward / ZeRO-3 (#7780)

tohtana committed 22 days ago

Verified 311674ff

docs: update README.md (#7781)

eltociear committed 22 days ago

Verified 5f2f9bf3

fix: skip compressed allreduce for empty tensors (#7769)

T1mn committed 23 days ago

Verified 97e74301

Fix multiprocessing testcase (#7743)

k-artem committed 23 days ago

Verified 7bbb6c77

Older