microsoft/DeepSpeed

Pull Requests Commits

Increase coverage for HPU

loadams committed 1 year ago

fdb02ea3

Added Gaudi2 CI tests. (#5275)

vshekhawat-hlab committed 1 year ago

Verified 330d36bb

Docs typos fix and grammar suggestions (#5322)

Gr0g0 committed 1 year ago

Verified cea5ea1e

Modified regular expression (#5306)

igeni committed 1 year ago

Verified a32789b0

fix comms dtype (#5297)

mayank31398 committed 1 year ago

Verified ed8aed57

Enable mixtral 8x7b autotp (#5257)

Yejing-Lai committed 1 year ago

Verified 8d98e171

support bf16_optimizer moe expert parallel training and moe EP grad_scale/grad_norm fix (#5259)

inkcherry committed 1 year ago

Verified e5dd5501

Fixed Accelerate Link (#5314)

wkaisertexas committed 1 year ago

Verified 4520edd6

Add intel gaudi to contributed HW in README (#5300)

BacharL committed 1 year ago

Verified 0529eac6

Add Habana Gaudi2 CI badge to the README (#5286)

loadams committed 1 year ago

Verified 3dd3d514

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

Older