huggingface/transformers

Pull Requests Commits

low cpu usage default to true

ArthurZucker committed 1 year ago

d58107a3

[tests] fix deepspeed zero3 config for `test_stage3_nvme_offload` (#31881)

faaany committed 1 year ago

Verified 25e5e3fa

Speedup model init on CPU (by 10x+ for llama-3-8B as one example) (#31771)

muellerzr committed 1 year ago

Verified e0dfd7bc

Cambricon MLUs support SDPA and flash_attn (#31102)

huismiling committed 1 year ago

Verified 03a3becc

Fix the incorrect permutation of gguf (#31788)

PenutChen committed 1 year ago

Verified ac946aac

Generate: doc nits (#31982)

gante committed 1 year ago

Verified 6fbea6d2

Masking: remove flakiness from test (#31939)

gante committed 1 year ago

Verified e4682de6

Avoid race condition (#31973)

ydshieh committed 1 year ago

Verified a1a34657

Notify new docker images built for circleci (#31701)

ydshieh committed 1 year ago

Verified 11efb4fc

fix: Fixed the arguments in `create_repo()` function call (#31947)

Sai-Suraj-27 committed 1 year ago

Verified 556a4205

Generate: handle `logits_warper` update in models with custom generate fn (#31957)

gante committed 1 year ago

Verified 90750042

fix: Removed a wrong key-word argument in `sigmoid_focal_loss()` function call (#31951)

Sai-Suraj-27 committed 1 year ago

Verified 454bc14d

Whisper: move to tensor cpu before converting to np array at decode time (#31954)

gante committed 1 year ago

Verified a5c642fe

Generate: v4.42 deprecations 🧹🧹 (#31956)

gante committed 1 year ago

Verified df1c248a

Generate: remove deprecated code due to `Cache` and `cache_position` being default (#31898)

gante committed 1 year ago

Verified 739a6316

Fix `GenerationMixin.generate` compatibility with pytorch profiler (#31935)

fxmarty committed 1 year ago

Verified 8480fda6

fix prompt strip to support tensors and np arrays (#27818)

AvivSham committed 1 year ago

Verified 7f79a973

Docker: TF pin on the consistency job (#31928)

gante committed 1 year ago

Verified d1a1bcf5

[Bug Fix] fix qa pipeline tensor to numpy (#31585)

jiqing-feng committed 1 year ago

Verified aec1ca3a

Adding hiera (#30356)

Namangarg110 committed 1 year ago

Verified c1e139c2

Allow `Trainer.get_optimizer_cls_and_kwargs` to be overridden (#31875)

apoorvkh committed 1 year ago

Verified 574e68d5

🚨 fix(SigLip): remove spurious exclusion of first vision output token (#30952)

transmissions11 committed 1 year ago

Verified 52585019

Generate: fix `SlidingWindowCache.reset()` (#31917)

gante committed 1 year ago

Verified 6a05f68f

Refactor flash attention implementation in transformers (#31446)

ArthurZucker committed 1 year ago

Verified e3143952

Fix fx tests with inputs_embeds (#31862)

fxmarty committed 1 year ago

Verified ad4ef3a2

Add warning message for beta and gamma parameters (#31654)

OmarManzoor committed 1 year ago

Verified 1499a550

add gather_use_object arguments II (#31799)

SangbumChoi committed 1 year ago

Verified 23d6d0cc

fix: Fixed the `1st argument` name in classmethods (#31907)

Sai-Suraj-27 committed 1 year ago

Verified 2e48b3e8

Fix missing methods for Fuyu (#31880)

Isotr0py committed 1 year ago

Verified 48c20700

[`Gemma2`] Support FA2 softcapping (#31887)

ArthurZucker committed 1 year ago

Verified f4ec7a28

Older