NielsRogge/transformers

Pull Requests Commits

Remove no_split_modules

NielsRogge committed 2 years ago

495c87b5

NielsRogge committed 2 years ago

d0fe6262

NielsRogge committed 2 years ago

ceac163a

feat: use `warning_advice` for tensorflow warning (#29540)

winstxnhdw committed 2 years ago

Verified 0290ec19

Fix eval thread fork bomb (#29538)

muellerzr committed 2 years ago

Verified 469c1328

[tests] use the correct `n_gpu` in `TrainerIntegrationTest::test_train_and_eval_dataloaders` for XPU (#29307)

faaany committed 2 years ago

Verified 3f6973db

Fix WhisperNoSpeechDetection when input is full silence (#29065)

ylacombe committed 2 years ago

Verified 1ba89dc2

fix typos in FSDP config parsing logic in `TrainingArguments` (#29189)

yundai424 committed 2 years ago

Verified 697f05ba

Make sliding window size inclusive in eager attention (#29519)

jonatanklosko committed 2 years ago

Verified 608fa549

StableLM: Fix dropout argument type error (#29236)

liangjs committed 2 years ago

Verified f386c51a

[tests] use `torch_device` instead of `auto` for model testing (#29531)

faaany committed 2 years ago

Verified 1ea3ad1a

Typo fix in error message (#29535)

clefourrier committed 2 years ago

Verified 14536c33

fix image-to-text batch incorrect output issue (#29342)

sywangyi committed 2 years ago

Verified 8ee1d472

[tests] add the missing `require_sacremoses` decorator (#29504)

faaany committed 2 years ago

Verified 8e589c83

Generate: left-padding test, revisited (#29515)

gante committed 2 years ago

Verified bc764f42

Typo in mlx tensor support (#29509)

pcuenca committed 2 years ago

Verified 631fa7bf

Fix `VisionEncoderDecoder` Positional Arg (#29497)

nickthegroot committed 2 years ago

Verified b338a6c3

Set `inputs` as kwarg in `TextClassificationPipeline` (#29495)

alvarobartt committed 2 years ago

Verified ddf177ee

test_generation_config_is_loaded_with_model - fall back to pytorch model for now (#29521)

amyeroberts committed 2 years ago

Verified 4ed9ae62

Add support for metadata format MLX (#29335)

alexweberk committed 2 years ago

Verified 45c06510

Flava multimodal add attention mask (#29446)

zucchini-nlp committed 2 years ago

Verified 923733c2

fix: Avoid error when fsdp_config is missing xla_fsdp_v2 (#29480)

ashokponkumar committed 2 years ago

Verified 9288e759

Revert "Automatic safetensors conversion when lacking these files (#2… (#29507)

LysandreJik committed 2 years ago

Verified f6133d76

v4.39 deprecations 🧼 (#29492)

gante committed 2 years ago

Verified ffe60fdc

Enable BLIP for auto VQA (#29499)

regisss committed 2 years ago

Verified 979fccc9

Fix: Disable torch.autocast in RotaryEmbedding of Gemma and LLaMa for MPS device (#29439)

currybab committed 2 years ago

Verified d45f47ab

Substantially reduce memory usage in _update_causal_mask for large batches by using .expand instead of .repeat [needs tests+sanity check] (#29413)

nqgl committed 2 years ago

Verified 2a939f20

Fix `TextGenerationPipeline.__call__` docstring (#29491)

alvarobartt committed 2 years ago

Verified 965cf677

added the max_matching_ngram_size to GenerationConfig (#29131)

mosheber committed 2 years ago

Verified 19fb1e22

Generate: torch.compile-ready generation config preparation (#29443)

gante committed 2 years ago

Verified ddb4fda3

Older