microsoft/DeepSpeed

Pull Requests Commits

Add Muon Optimizer Support to ZeRO3

Zhipeng Wang committed 1 day ago

f8998a86

Fix AutoTP test numerical tolerance with rtol (#7794)

tohtana committed 1 day ago

Verified 374f6d09

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

aeeeeeep committed 1 day ago

Verified 60876090

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

tohtana committed 1 day ago

Verified 991ebd72

Fix Evoformer compilation (#7760)

sdvillal committed 3 days ago

Verified 43125a75

Fix Ulysses PEFT test (#7784)

tohtana committed 3 days ago

Verified 491a38ce

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

tohtana committed 4 days ago

Verified 311674ff

docs: update README.md (#7781)

eltociear committed 4 days ago

Verified 5f2f9bf3

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

T1mn committed 5 days ago

Verified 97e74301

Fix multiprocessing testcase (#7743)

k-artem committed 6 days ago

Verified 7bbb6c77

Update CLA to refer to DCO (#7778)

loadams committed 6 days ago

Verified e31daa8c

[Docs] Fix `docs/_pages/config-json.md` format (#7779)

ooooo-create committed 6 days ago

Verified 79af89be

Remove cron/PR triggers for outdated V100 tests (#7777)

loadams committed 7 days ago

Verified 53c87ce8

Add timeout to test workflows (#7774)

tohtana committed 8 days ago

Verified 52361686

Update README with newer status badges for CI

loadams committed 8 days ago

b80f453f

fix: update Megatron-DeepSpeed tutorial to match current repo structure (#7761)

nathon-lee committed 10 days ago

Verified 8a9369d0

fix: delete using namespace std. (#7766)

nathon-lee committed 11 days ago

Verified 816e4aef

Udpate workflow trigger (#7768)

tohtana committed 11 days ago

Verified d988f6ca

Various fixes to run on mps (#7767)

jeffra committed 11 days ago

Verified e5b93809

Update version.txt after 0.18.4 release (#7765)

loadams committed 13 days ago

Verified 6b878203

fix: avoid IndexError in BF16_Optimizer.destroy() when using DummyOptim (#7763)

nathon-lee committed 14 days ago

Verified b35d9eb0

[AMD] Skipped amp testcases for ROCm (#7745)

k-artem committed 15 days ago

Verified d7bb6f34

Fix DeepCompile for PyTorch 2.8/2.9 compatibility (#7755)

tohtana committed 18 days ago

Verified 2390bc5c

Add fallback to BF16 support check (#7754)

tohtana committed 18 days ago

Verified 1d140f80

Add CI workflow to run tests on AWS (#7753)

tohtana committed 18 days ago

Verified bfb66c65

Enable python 3.11 and 3.12 tests (#7007)

loadams committed 20 days ago

Verified c0e9b2c9

fix: remove premature MPI environment variable check in OpenMPIRunner (#7751)

nathon-lee committed 20 days ago

Verified 4809072a

Fix OnebitLamb NaN propagation with empty parameters (#7736)

Rakshit-gen committed 27 days ago

Verified b4e74a91

Fix DecoupledCheckpointEngine deadlock and improve reliability (#7742)

Rakshit-gen committed 29 days ago

Verified ba29fdca

Fix Nebula checkpoint engine commit() API mismatch (#7740)

Rakshit-gen committed 30 days ago

Verified 6d9c3dc0

Older