microsoft/DeepSpeed

Pull Requests Commits

Merge branch 'master' into loadams/pynvml

loadams committed 1 year ago

Verified 2a2a3013

enable phi2 autotp (#5436)

Yejing-Lai committed 1 year ago

Verified 3a7f3aa8

Merge branch 'master' into loadams/pynvml

loadams committed 1 year ago

Verified a41137a1

Formatting fixes

loadams committed 1 year ago

bacd50aa

Switch pynvml to nvidia-ml-py

loadams committed 1 year ago

7236186f

[manifest] update mainfest to add hpp file in csrc. (#5522)

ys950902 committed 1 year ago

Verified 4696afd2

Fused adam for HPU (#5500)

BacharL committed 1 year ago

Verified df4ef0ab

enable phi3_mini autotp (#5501)

Yejing-Lai committed 1 year ago

Verified 3dd7ccff

Fix compile wrapper (#5455)

BacharL committed 1 year ago

Verified 0b224edc

Fix crash when creating Torch tensor on NPU with device=get_accelerator().current_device() (#5464)

harygo2 committed 1 year ago

Verified 0fc19b6a

re-introduce: stage3: efficient compute of scaled_global_grad_norm (#5493)

nelyahu committed 1 year ago

Verified 90793aab

Un-pin torch version in nv-torch-latest back to latest and skip test_compile_zero tests on v100 (#5459)

loadams committed 1 year ago

Verified f32ad3e1

fix: swapping order of parameters in create_dir_symlink method. (#5465)

alvieirajr committed 1 year ago

Verified 059bb208

Update with ops not supported on Windows (#5468)

loadams committed 1 year ago

Verified 4c15ad9f

Update ds-chat CI workflow paths to include zero stage 1-3 files (#5462)

lekurile committed 1 year ago

Verified e37296b2

Revert "stage3: efficient compute of scaled_global_grad_norm (#5256)" (#5461)

lekurile committed 1 year ago

Verified bc48371c

Fix torch.compile error for PyTorch v2.3 (#5463)

tohtana committed 1 year ago

Verified fcc731f0

Add getter and setter methods for compile_backend across accelerators. (#5299)

vshekhawat-hlab committed 1 year ago

Verified fa8458b1

Update version.txt after 0.14.2 release (#5458)

mrwyattii committed 1 year ago

Verified fbdf0eaf

Update PyTest torch version to match PyTorch latest official (2.3.0) (#5454)

loadams committed 1 year ago

Verified 5f631abc

Update README.md (#5453)

Jhonso7393 committed 1 year ago

Verified ad202795

OptimizedLinear implementation (#5355)

jeffra committed 1 year ago

Verified 5e6c9b93

set the default to use set_to_none for clearing gradients in BF16 optimizer. (#5434)

inkcherry committed 1 year ago

Verified c66bc426

Improve parallel process of universal checkpoint conversion (#5343)

tohtana committed 1 year ago

Verified c292b03a

64bit indexing fused adam (#5187)

garrett4wade committed 1 year ago

Verified 9b6ef9e1

add device config env for the accelerator (#5396)

shiyuan680 committed 1 year ago

Verified 3f875d95

Fix sorting of shard optimizer states files for universal checkpoint (#5395)

tohtana committed 1 year ago

Verified 99951caa

Selective dequantization (#5375)

RezaYazdaniAminabadi committed 1 year ago

Verified c632ea09

Parallel map step for `DistributedDataAnalyzer` map-reduce (#5291)

bm-synth committed 1 year ago

Verified 64defe65

Bug fix in `split_index` method (#5292)

bm-synth committed 1 year ago

Verified aaaf8bc5

Older