huggingface/transformers

Pull Requests Commits

Avoid unnecessary device operations in loss computing (#36950)

cyyever committed 1 year ago

Verified 8c5e29ba

clean pipeline question_answering. (#36986)

湛露先生 committed 1 year ago

Verified 471cf1de

[generate, cache] handle more complex device maps (#37014)

gante committed 1 year ago

Verified 29f322d0

[audio utils] fix fft_bin_width computation (#36603)

eustlb committed 1 year ago

Verified fb8e6c50

[chat templates} support loading audio from video (#36955)

zucchini-nlp committed 1 year ago

Verified e97c7600

Fixup for distill_any_depth conversion script (#37043)

qubvel committed 1 year ago

Verified c7bc79bd

Optimize `to_py_obj` for python-native numeric lists and scalars (#36885)

n0gu-furiosa committed 1 year ago

Verified d1eafe8d

fix pegasus init weights and other copied models (#36844)

jiqing-feng committed 1 year ago

Verified 0e56fb69

Add Distill Any Depth (#36614)

keetrap committed 1 year ago

Verified 7e813f9c

Skip FP8 linear tests For device capability < 9.0(#37008)

MekkCyber committed 1 year ago

Verified 92429057

remove redundant code in trainer (#36994)

hiyouga committed 1 year ago

Verified 279c2e30

Mark 2 tests as flaky for now (#37038)

ydshieh committed 1 year ago

Verified d13c390d

[Modeling] Load FP8 safetensors such as DeepSeek (#36828)

kylesayrs committed 1 year ago

Verified d6d930a6

Fix PixtralProcessor patch_size when spatial_merge_size is used (#37019)

mgoin committed 1 year ago

Verified 927ce1d3

Support QuestionAnswering Module for ModernBert based models. (#35566)

bakrianoo committed 1 year ago

Verified 49b5ab6a

fix transformers_cli import relative path issue (#36989)

yao-matrix committed 1 year ago

Verified 5b08db88

[docs] Attention mask image (#36970)

stevhliu committed 1 year ago

Verified 3a8ec8c4

Remove deprecated training arguments (#36946)

cyyever committed 1 year ago

Verified 2b550c47

fix typos in the code comments and error messages (#36993)

threewebcode committed 1 year ago

Verified 44715225

Log the correct learning rate (#36973)

SunMarc committed 1 year ago

Verified 79d6f9fd

Fix device_map check for ggml files (#37003)

MekkCyber committed 1 year ago

Verified 13d36e89

Fix removing "cpu" from frozenset in bitsandbytes.py to allow better ROCm support. (#36975)

anadon committed 1 year ago

Verified 021006e1

Allow easy registration of custom attention functions (#36889)

Cyrilvallez committed 1 year ago

Verified 788e1092

Fix get_device_properties (#36997)

ivarflakstad committed 1 year ago

Verified ad5d40de

Fix Optional type annotation (#36841)

cyyever committed 1 year ago

Verified 8084b262

Install `networkx==3.2.1` manually in some CircleCI jobs after #36957 (#37000)

ydshieh committed 1 year ago

Verified b56d8f07

Use torch.expm1 (#36995)

cyyever committed 1 year ago

Verified 78afa1c5

byebye CircleCI TF jobs (#36998)

ydshieh committed 1 year ago

Verified 181d4530

Fix tensor dtype mismatch (#36985)

cyyever committed 1 year ago

Verified e7139d06

🚨Deprecate legacy argument for image-text-to-text models and adopt new behavior by default (#36307)

yonigozlan committed 1 year ago

Verified be37d34f

Newer Older