microsoft/DeepSpeed

Pull Requests Commits

stas00 committed 100 days ago

3870f5c7

logging: Also set log level of logger handlers

eternalNight committed 101 days ago

6d418802

Update version.txt post 0.17.6 release (#7572)

loadams committed 103 days ago

Verified 80033a82

[bugfix] fix partition context unpatch (#7566)

hjh0119 committed 104 days ago

Verified e4f6da96

scripts: Check .is_cuda only in non-C++ files (#7561)

eternalNight committed 104 days ago

Verified 6b731c5c

Make Muon optimizer easier to enable (#7555)

delock committed 105 days ago

Verified 2585881a

fix npu device_id AttributeError issue (#7560)

we1sper committed 106 days ago

Verified aa539c6d

deepcompile: Create a full list of no-copy ops (#7562)

eternalNight committed 106 days ago

Verified 2d84be81

deepcompile: Record graph order using OrderedDict (#7563)

eternalNight committed 107 days ago

Verified e9d5d416

deepcompile: Create dummy inputs using empty_strided (#7564)

eternalNight committed 107 days ago

Verified 660ee895

Add dependency for deepcompile test (#7558)

tohtana committed 110 days ago

Verified d40a0f5d

Move modal tests to tests/v1 (#7557)

tohtana committed 110 days ago

Verified b9bd03a2

Fix gradient buffer access for DeepCompile Z1/2 (#7548)

tohtana committed 112 days ago

Verified 0e859aa0

Limit random seed range in tests (#7553)

tohtana committed 112 days ago

Verified 0012ff6e

[MoE] Fix misuse of num_experts as expert parallel group size (ep_size) (#7551)

Flakes342 committed 113 days ago

Verified 8cbbbb53

[alstn tutorial] support bs>1 (#7550)

sfc-gh-sbekman committed 113 days ago

Verified 533e834b

Revert "Add index to HPU devices (#7497)" (#7545)

deepcharm committed 114 days ago

Verified 450b965e

Improve error message and reduce validation in autocast test (#7547)

tohtana committed 118 days ago

Verified b82ef716

avoid setting device_id to `init_process_group` (#7542)

kaixuanliu committed 118 days ago

Verified 08879a39

fix get_cuda_compile_flag (#7521)

mingjielu committed 118 days ago

Verified 78a74874

Autotune ZenFlow affinity (#7506)

delock committed 119 days ago

Verified 43537d0a

Relax restrictions of torch.autocast integration (#7543)

tohtana committed 119 days ago

Verified 66bf2a64

Fix zenflow_torch_adam.py (#7544)

stas00 committed 119 days ago

Verified 8af75487

Fix scaling and allgather with `torch.autocast` (#7534)

tohtana committed 120 days ago

Verified 1e183a6a

fixed DeepSpeedCPULion with ZeRO-Offload bug (#7531)

qibin0506 committed 120 days ago

Verified c07b3abf

docs typo: `lrrt.md`, reference to `cycle_min_lr` should be `cycle_max_lr` (#7530)

jakehemmerle committed 120 days ago

Verified 4d83f3fe

[doc] fixing moe tutorial (#7538)

stas00 committed 120 days ago

Verified 9e4957eb

[logging] less startup noise (#7526)

stas00 committed 120 days ago

Verified 066d9120

undo the revert (#7536)

stas00 committed 120 days ago

Verified 411e20a3

fix typo s/1014 /1024 (#7528)

digger-yu committed 122 days ago

Verified 902e78c9

Older