microsoft/DeepSpeed

Pull Requests Commits

Add Habana Gaudi2 CI badge to the README

loadams committed 1 year ago

013c4d94

Fix loading a universal checkpoint (#5263)

tohtana committed 1 year ago

Verified b112c99e

Add HPU accelerator support in unit tests. (#5162)

vshekhawat-hlab committed 1 year ago

Verified 2df8e234

Average only valid part of the ipg buffer. (#5268)

BacharL committed 1 year ago

Verified 75e579e7

Added HPU backend support for torch.compile tests. (#5269)

vshekhawat-hlab committed 1 year ago

Verified a6fb4d3e

Enabled LMCorrectness inference tests on HPU. (#5271)

vshekhawat-hlab committed 1 year ago

Verified cf58c535

Skip 1Bit Compression and sparsegrad tests for HPU. (#5270)

vshekhawat-hlab committed 1 year ago

Verified d2ef20e7

Fix attention mask handling in the Hybrid Engine Bloom flow (#5101)

deepcharm committed 1 year ago

Verified d9e12d3a

Add CI for Habana Labs HPU/Gaudi2 (#5244)

loadams committed 1 year ago

Verified 2989cf77

Replace HIP_PLATFORM_HCC with HIP_PLATFORM_AMD (#5264)

rraminen committed 1 year ago

Verified 49359d0b

Make op builder detection adapt to accelerator change (#5206)

delock committed 1 year ago

Verified c08e69f2

Set tp world size to 1 in ckpt load, if MPU is not provided (#5243)

samadejacobs committed 1 year ago

Verified 535a908f

Add contributed HW support into README (#5240)

delock committed 1 year ago

Verified 74910a97

Fp6 blog chinese (#5239)

xiaoxiawu-microsoft committed 1 year ago

Verified d1536e44

Update version.txt after 0.14.0 release (#5238)

mrwyattii committed 1 year ago

Verified 5e883318

Update version.txt

loadams committed 1 year ago

ce78a632

FP6 blog (#5235)

loadams committed 1 year ago

Verified 0a979f8b

FP6 quantization end-to-end. (#5234)

loadams committed 1 year ago

Verified ccfdb84e

MOE gate fixes and enhancements (#5156)

mosheisland committed 1 year ago

Verified 5a2e705b

Update version.txt after 0.13.5 release (#5229)

mrwyattii committed 1 year ago

Verified db70c183

fix fused_qkv model accuracy issue (#5217)

Yejing-Lai committed 1 year ago

Verified bc0d2465

Use `deepspeed.comm` instead of `torch.distributed` (#5225)

jinyouzhi committed 1 year ago

Verified acf07398

Fix moe cpu offload (#5220)

RezaYazdaniAminabadi committed 1 year ago

Verified e6e8c137

Rename nv-torch-latest-cpu workflow to cpu-torch-latest (#5226)

loadams committed 1 year ago

Verified 3e06a154

Add fp16 support of Qwen1.5 models (0.5B to 72B) to DeepSpeed-FastGen (#5219)

ZonePG committed 1 year ago

Verified bcc617a0

[zero++] Synchronize at the end of secondary partitioning and simplify the logic (#5216)

ByronHsu committed 1 year ago

Verified 4578c249

Allow specifying MII branch on MII CI (#5208)

mrwyattii committed 1 year ago

Verified 857584fd

Fix assertion to run pipeline engine with a compiled module (#5197)

tohtana committed 1 year ago

Verified aed599b4

Update version.txt after 0.13.4 release (#5196)

mrwyattii committed 1 year ago

Verified 07a1ede4

Add script to check for `--extra-index-url` (#5184)

loadams committed 1 year ago

Verified 5115df3f

Older