huggingface/accelerate

Pull Requests Commits

Release: v1.4.0

muellerzr committed 1 year ago

b431d1f0

Torchao float8 training (#3348)

muellerzr committed 1 year ago

Verified 8039158d

enable xpu (#3397)

faaany committed 1 year ago

Verified e34db4d0

[`memory leak`] Replace GradientState -> DataLoader reference with weakrefs (#3391)

tomaarsen committed 1 year ago

Verified 526925b4

[examples] upgrade code for seed setting (#3387)

faaany committed 1 year ago

Verified 24f8d027

fix: typos in documentation files (#3388)

maximevtush committed 1 year ago

Verified 5cc99e6e

works for fp8 with deepspeed (#3361)

XiaobingSuper committed 1 year ago

Verified ce636234

fix torch_dtype in estimate memory (#3383)

SunMarc committed 1 year ago

Verified f19b9570

[tests] Fix bnb cpu error (#3351)

faaany committed 1 year ago

Verified 81d8a035

deepspeed github repo move (#3376)

stas00 committed 1 year ago

Verified f0764955

feat: support tensor parallel & Data loader (#3173)

kmehant committed 1 year ago

Verified 03153658

[tests] enable more bnb tests on XPU (#3350)

faaany committed 1 year ago

Verified 675e35bc

Support more functionalities for MUSA backend (#3359)

fmo-mt committed 1 year ago

Verified 8f2d31c5

[tests] remove `require_non_xpu` test markers (#3301)

faaany committed 1 year ago

Verified 4c2c89ea

[tests] make cuda-only test work on other hardware accelerators (#3302)

faaany committed 1 year ago

Verified 28c171b0

[Dev] Update release directions (#3352)

muellerzr committed 1 year ago

Verified 65356780

muellerzr committed 1 year ago

78b8126b

[tests] enable BNB test cases in `tests/test_quantization.py` on XPU (#3349)

faaany committed 1 year ago

Verified 7e324103

fix triton version check (#3345)

faaany committed 1 year ago

Verified 02d25612

dataloader: check that in_order is in kwargs before trying to drop it (#3346)

dvrogozh committed 1 year ago

Verified fbfa53bc

[docs] fix typo, change "backoff_filter" to "backoff_factor" (#3296)

suchot committed 1 year ago

Verified d09040df

add torchdata version check to avoid "in_order" error (#3344)

faaany committed 1 year ago

Verified 828aae4e

Fix for offloading when using TorchAO >= 0.7.0 (#3332)

a-r-r-o-w committed 1 year ago

Verified f0b03055

latest bnb no longer has optim_args attribute on optimizer (#3311)

winglian committed 1 year ago

Verified 80973430

[tests] make cuda-only test case device-agnostic (#3340)

faaany committed 1 year ago

Verified c67d47ae

Fix offload generate tests (#3334)

SunMarc committed 1 year ago

Verified 8c423cff

feat(tpu): remove nprocs from xla.spawn (#3324)

tengomucho committed 1 year ago

Verified 95f34d62

Fixup docker build err (#3333)

muellerzr committed 1 year ago

Verified ba90f856

Bye bye torch <2 (#3331)

muellerzr committed 1 year ago

Verified b13aadcb

Ensure that tied parameter is children of module (#3327)

pablomlago committed 1 year ago

Verified 58f14364

Older