microsoft/DeepSpeed

Pull Requests Commits

Fix hook count performance regression from v0.18.5 (#7886)

tohtana committed 2 days ago

Verified 6c59d544

Add document section explaining autocast nesting (#7883)

tohtana committed 3 days ago

Verified 4dba1e20

Remove amp() from abstract accelerator (#7879)

delock committed 5 days ago

Verified a41a96b1

XPU use stock pytorch instead of Intel Extension for PyTorch (#7877)

delock committed 5 days ago

Verified d8e15da4

Fix import deepspeed crash on PyTorch v2.3 + Python 3.12 (#7875)

tohtana committed 5 days ago

Verified 04d69cc6

Fix no-grad grad-fn lookup in ZeRO hook counting on PyTorch 2.3 (#7830) (#7841)

tohtana committed 6 days ago

Verified bffaf457

Fix: only add parameter with grads to parameter group (#7869)

delock committed 6 days ago

Verified 116dbe28

perf: use deque for FIFO queues in sequence parallel, superoffload, and compile (#7880)

giulio-leone committed 6 days ago

Verified a15e557d

fix: replace 35 bare except clauses with except Exception (#7873)

haosenwang1018 committed 9 days ago

Verified ae21699a

Fix broken links and add AutoTP Training tutorial to sidebar nav (#7874)

tohtana committed 9 days ago

Verified efc0b49a

Schedule nightly full test (#7870)

tohtana committed 11 days ago

Verified 0416cf68

Fix Zero-3 static scale assertion in fp16 test (#7866)

tohtana committed 13 days ago

Verified 93524c89

Throw error when parameter is modified in GatheredParameters (#7832)

tohtana committed 13 days ago

Verified 57b10d54

Fix compilation of Evoformer (#7862)

Flamefire committed 14 days ago

Verified c9e652cf

Fix ROCm BF16 conversion intrinsics in inference v2 (#7843) (#7846)

tohtana committed 16 days ago

Verified dbc1b075

Add EXAONE 4.0 model support for Inference V2 (#7853)

Bias92 committed 18 days ago

Verified f3a9819c

Add news entry for DeepSpeed updates (#7854)

PKUWZP committed 19 days ago

Verified 8a63a2a9

Enable shm_comm support for arm (#7800)

phalani-paladugu committed 20 days ago

Verified 7f49367a

Z1/2 init: flatten params on device (#7828)

ksugama committed 22 days ago

Verified 84af8221

Merge pull request #7850 from deepspeedai/loadams/update-post-0.18.6

loadams committed 22 days ago

Verified b74df164

Replace torch.jit.script with torch.compile (#7835) (#7840)

tohtana committed 23 days ago

Verified d2ca6e75

Fix bf16 gradient norm divergence with ZeRO stage 0 (#7839)

tohtana committed 23 days ago

Verified 1752c2ab

fix: replace deprecated fractions.gcd with math.gcd (#7845)

Mr-Neutr0n committed 23 days ago

Verified 2c362837

Support new python 3.14 annotation handling (#7831)

sdvillal committed 25 days ago

Verified 72b9465c

Fix AutoTP custom patterns: respect use_default_specs (#7827)

tohtana committed 28 days ago

Verified a44fb58f

Fix gradient is ready with z2 (#7829)

sfc-gh-truwase committed 31 days ago

Verified 0ccb2bb6

Support custom partitioning patterns for AutoTP (#7806)

tohtana committed 35 days ago

Verified 6b9cab1d

Skip sequence parallel operations during eval (#7821)

jp1924 committed 35 days ago

Verified a972f7c6

Fix leaf module race condition (#7825)

tohtana committed 35 days ago

Verified 52b1d4d6

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

loadams committed 36 days ago

Verified 7b15a6a7

Older