microsoft/DeepSpeed

Pull Requests Commits

hwchen2017 committed 2 hours ago

6aaaf8c1

Support a suffix

hwchen2017 committed 4 hours ago

c59029ea

Unify compatibility check

hwchen2017 committed 7 hours ago

3c55eb0d

hwchen2017 committed 7 hours ago

4f53ff0f

remove hardcoded cpp standard

hwchen2017 committed 7 hours ago

4a67eaff

Fix problem in prebuilt ops

hwchen2017 committed 7 hours ago

eea2048d

Fix cuda mismatch problem

hwchen2017 committed 7 hours ago

651fe0a5

move cuda/rocm related code to cudaopbuilder

hwchen2017 committed 7 hours ago

c7f4a24e

remove build for cpu

hwchen2017 committed 7 hours ago

9b2a0601

Treat an empty subset selection as success in modal-torch-latest (#9053)

GuCarpenter committed 16 hours ago

Verified b9100c54

Size HybridEngine rollout KV caches from local prefill shapes under AutoTP (#9008)

jinyouzhi committed 17 hours ago

Verified 2dabbbe3

Synchronize ZeRO-3 rollout early stopping (#8390)

LiRunGuo committed 22 hours ago

Verified 8a6b53fe

ZeRO-2 CPU offload: accumulate in the compact partition between requests (#9044)

sfc-gh-truwase committed 1 day ago

Verified f9fe4df3

Update an affiliation in COMMITTERS.md (#9050)

tohtana committed 1 day ago

Verified 51d86659

[NPU] Fix the fused LAMB step and inference int4/int8 dequantization on Ascend (#9043)

VenusTZZ committed 1 day ago

Verified f7a71777

Use C++20 for CUDA/ROCm op builders on torch 2.12+ (#9041)

VenusTZZ committed 1 day ago

Verified cbed412d

Code cleaning: remove commented out debug prints and code (#9036)

hwchen2017 committed 2 days ago

Verified 6f96b4a0

Fix memory leak in engine and zero optimizer (#9034)

hwchen2017 committed 2 days ago

Verified f32e164b

Persist ZeRO-3 HybridMegatron QKV layout changes (#8392)

LiRunGuo committed 2 days ago

Verified 988c88a0

ZeRO-1/2: one CPU-offload accumulator per group; zero idle gradients (#9018)

sfc-gh-truwase committed 2 days ago

Verified a4c989b7

[VocabTP] Add vocabulary-parallel generalized JSD with Triton kernels (#8958)

jinyouzhi committed 3 days ago

Verified da79d744

feat(rollout): non-greedy sampling on hybrid engine graph decode path (#8939)

anyadontfly committed 3 days ago

Verified 02f79784

Treat an all-deselected selection as success in cpu-torch-latest-v1-multirank (#9032)

delock committed 3 days ago

Verified 57ec847d

Fix routed-expert gradient clipping for AutoEP + AutoTP folding (#8823)

0z5a committed 4 days ago

Verified a7ebdf96

ZeRO-3 NVMe offload: fix "param already assigned swap buffer" in the parameter prefetch (#9014)

xylian86 committed 4 days ago

Verified 6c7a6e0e

fix test_multiple_request_boundaries_before_step (#9028)

stas00 committed 4 days ago

Verified af519898

Keep fp32 CPU optimizer states in fp32 when loading a state_dict (#8974)

xylian86 committed 4 days ago

Verified b43c383f

DeepNVMe: fix silent I/O failures on misaligned buffers (#8985)

xylian86 committed 4 days ago

Verified 1b022dcd

CI: stop pip check from failing the AWS fallback on deepspeed-kernels (#9020)

pengdurice committed 4 days ago

Verified 21e7ba48

AutoEP: read and write the transformers output-capture registry as a dict (#9029)

pengdurice committed 4 days ago

Verified e550bedb

Older