microsoft/DeepSpeed

Pull Requests Commits

Merge branch 'master' into guanhua/overflow-check

tjruwase committed 1 year ago

Verified 7bd6aaff

GuanhuaWang committed 1 year ago

fda7b77f

Update _sidebar.scss (#5293)

fasterinnerlooper committed 1 year ago

Verified a9cbd688

Update engine.py to avoid torch warning (#5408)

etiennebonnafoux committed 1 year ago

Verified bc0f7747

Update 'create-pr' action in release workflow to latest (#5415)

loadams committed 1 year ago

Verified 34969d20

Fix a convergence issues in TP topology caused by incorrect grad_norm. (#5411)

inkcherry committed 1 year ago

Verified 0896503e

[XPU] Use non_daemonic_proc by default on XPU device (#5412)

ys950902 committed 1 year ago

Verified 258e500f

Remove dtype(fp16) condition check for residual_add unit test (#5329)

raza-sikander committed 1 year ago

Verified 89491053

Update version.txt after 0.14.1 release (#5413)

mrwyattii committed 1 year ago

Verified 2b3d31f9

Fix the FP6 kernels compilation problem on non-Ampere GPUs. (#5333)

JamesTheZ committed 1 year ago

Verified e3d873a0

stage3: efficient compute of scaled_global_grad_norm (#5256)

nelyahu committed 1 year ago

Verified 54c06872

fix pagable h2d memcpy (#5301)

GuanhuaWang committed 1 year ago

Verified 7b5b0660

Removal of cuda hardcoded string with get_device function (#5351)

raza-sikander committed 1 year ago

Verified f69f8840

Add custom reshaping for universal checkpoint (#5390)

tohtana committed 1 year ago

Verified 2c51aba0

Optimize the fp-dequantizer to get high memory-BW utilization (#5373)

RezaYazdaniAminabadi committed 1 year ago

Verified a8b82153

Update README.md for intel XPU support (#5389)

Liangliang-Ma committed 1 year ago

Verified cc9e7b9c

Cleanup required_torch_version code and references. (#5370)

loadams committed 1 year ago

Verified 6dcced1d

Update checkout action on workflows on ubuntu 20.04 (#5387)

loadams committed 1 year ago

Verified 63029e8f

Revert "Update XPU workflow too"

loadams committed 1 year ago

62ab1810

Update XPU workflow too

loadams committed 1 year ago

eb216a83

Update path name on xpu-max1100.yml, add badge in README (#5386)

loadams committed 1 year ago

Verified 7e126d20

Add CI for Intel XPU/Max1100 (#5376)

Liangliang-Ma committed 1 year ago

Verified 87373889

Increase CI coverage for Gaudi2 accelerator. (#5358)

vshekhawat-hlab committed 1 year ago

Verified ca4bc66b

Update pytest and transformers with fixes for pytest>= 8.0.0 (#5164)

loadams committed 1 year ago

Verified 72bd275b

Support MoE for pipeline models (#5338)

mosheisland committed 1 year ago

Verified 08e0733e

Auto convert moe param groups (#5354)

jeffra committed 1 year ago

Verified 42a8eaa7

CPU SHM based inference_all_reduce improve (#5320)

delock committed 1 year ago

Verified 731fd682

FP [6,8,12] quantizer op (#5336)

jeffra committed 1 year ago

Verified 3fbd01cc

Update workflows that use cu116 to cu117 (#5361)

loadams committed 1 year ago

Verified 4621ba4c

Ensure capacity does not exceed number of tokens (#5353)

jeffra committed 1 year ago

Verified 5fbc3eee

Older