deepspeedai/DeepSpeed

Pull Requests Commits

RTD: remove native pin backend docs not in this PR

sfc-gh-truwase committed 70 days ago

99b30652

RTD: drop libaio/async_io from pin_memory op docs

sfc-gh-truwase committed 70 days ago

7db4b06d

RTD: document pin_memory op in Host Memory Pinning

sfc-gh-truwase committed 70 days ago

bc07cc9e

Trim shared behaviors from torch vs native pin comparison table

sfc-gh-truwase committed 71 days ago

98670e13

Note that torch-pinned buffers are visible to AIO/GDS

sfc-gh-truwase committed 71 days ago

1f64b281

Fix DeepNVMe bounce-buffer note for torch pin backend

sfc-gh-truwase committed 71 days ago

2bcd18e5

Document torch vs native pin backend differences in a table

sfc-gh-truwase committed 71 days ago

3ca9cc2d

Extend unmanaged gradient accumulation to ZeRO offload

sfc-gh-truwase committed 72 days ago

d3dcf9ad

Restore the custom DCO workflow (#8220)

tohtana committed 73 days ago

Verified 20328cf5

Update version.txt after 0.19.4 release (#8219)

loadams committed 73 days ago

Verified 14364cd1

Update version.txt pre-0.19.4 release (#8218)

loadams committed 73 days ago

Verified c4550314

Unmanaged gradient accumulation: ZeRO stage 3 support (#8217)

sfc-gh-truwase committed 73 days ago

Verified e75d1150

Fix ZeRO-3 all_reduce param fetch stride for padded parameters (#8158)

ebarkhordar committed 73 days ago

Verified 4e3acb1e

Unmanaged gradient accumulation: ZeRO stage 2 support (#8203)

sfc-gh-truwase committed 73 days ago

Verified 3c737a1c

Stop trim_mean sorting the caller's list in place (#8199)

vineethsaivs committed 73 days ago

Verified b8b44809

Share DeepNVMe pinned-tensor manager and route swap buffers through I/O handles (#8212)

sfc-gh-truwase committed 73 days ago

Verified e1d6b4fe

Make PipelineModule.set_checkpoint_interval actually change the interval (#8178)

vineethsaivs committed 74 days ago

Verified 87e4c8c8

Reduce redundant work in AutoEP token routing (#8209)

hwchen2017 committed 75 days ago

Verified 58aab420

Fix MLU_Accelerator conformance to the DeepSpeedAccelerator ABC (#8208)

ebarkhordar committed 75 days ago

Verified 437226bf

Consolidate LR scheduler tensor-LR tests into a contract suite (#8206)

delock committed 75 days ago

Verified 6498adc5

Fix DeepCompile ZeRO-3 gathered parameter ownership (#8157)

tohtana committed 75 days ago

Verified d8dd9e43

Wiring up offload_opt_states (#8186)

pengdurice committed 75 days ago

Verified 76928ff9

Add Triton grouped-GEMM for MoE experts on Ampere/Ada (#8180)

hwchen2017 committed 76 days ago

Verified df84f6d8

Preserve tensor learning rates in OneCycle (#8205)

tohtana committed 76 days ago

Verified f6a386a3

Remove redundant DCO workflow (#8188)

tohtana committed 76 days ago

Verified d2bd087c

Add keyword argument support to activation checkpointing (#8182)

zupengwang committed 76 days ago

Verified ebf1531e

MoE dispatch: fix silent Tutel + TP corruption, speed up native and Tutel paths (#8195)

jinyouzhi committed 76 days ago

Verified 619e3199

Add managed_gradient_accumulation for ZeRO stage 0/1 (#8184)

sfc-gh-truwase committed 76 days ago

Verified 7a31fe66

[AutoTP] Fix ZeRO-3 checkpoint consolidation to gather across TP and DP (#8168)

delock committed 76 days ago

Verified eec237ee

Preserve tensor learning rates across scheduler updates (#8202)

n33levo committed 76 days ago

Verified 3b641487

Older