NielsRogge/transformers

Pull Requests Commits

Address comments

NielsRogge committed 2 years ago

d5d8ab3c

NielsRogge committed 2 years ago

4953af8f

Depth Anything: update conversion script for V2 (#31522)

pcuenca committed 2 years ago

Verified 1082361a

Fix Wav2Vec2 Fairseq conversion (weight norm state dict keys) (#31714)

gau-nernst committed 2 years ago

Verified a8fa6fbb

Fix galore lr display with schedulers (#31710)

vasqu committed 2 years ago

Verified a01b033c

Allow FP16 or other precision inference for Pipelines (#31342)

aliencaocao committed 2 years ago

Verified ac262604

Repeating an important warning in the chat template docs (#31796)

Rocketknight1 committed 2 years ago

Verified e7868444

Add training support for SigLIP (#31495)

aliencaocao committed 2 years ago

Verified 1d3eaa6f

Code agent: allow function persistence between steps (#31769)

aymeric-roucher committed 2 years ago

Verified 15560252

Fix gemma tests (#31794)

ydshieh committed 2 years ago

Verified eef0507f

Update CometCallback to allow reusing of the running experiment (#31366)

Lothiraldan committed 2 years ago

Verified 9e599d1d

Exclude torch.compile time from metrics computation (#31443)

zxd1997066 committed 2 years ago

Verified d19b5a90

Make tensor device correct when ACCELERATE_TORCH_DEVICE is defined (#31751)

kiszk committed 2 years ago

Verified 2aa2a144

Fix serialization for offloaded model (#31727)

SunMarc committed 2 years ago

Verified 8c5c180d

Fix ClapProcessor to merge feature_extractor output into the returned BatchEncoding (#31767)

mxkopy committed 2 years ago

Verified eaa5f414

Add torch_empty_cache_steps to TrainingArguments (#31546)

aliencaocao committed 2 years ago

Verified 43ffb785

Fix Gemma2 types (#31779)

hiyouga committed 2 years ago

Verified cee768d9

`pytest_num_workers=4` for some CircleCI jobs (#31764)

ydshieh committed 2 years ago

Verified 87726a08

Fix RT-DETR weights initialization (#31724)

qubvel committed 2 years ago

Verified 048f599f

Fix RT-DETR cache for generate_anchors (#31671)

qubvel committed 2 years ago

Verified b9752161

[fix bug] logits's shape different from label's shape in preprocess_logits_for_metrics (#31447)

wiserxin committed 2 years ago

Verified 534cbf8a

Add ignore_errors=True to trainer.py rmtree in _inner_training_loop (#31668)

njbrake committed 2 years ago

Verified 65a02cd2

Gemma 2: Update slow tests (#31759)

gante committed 2 years ago

Verified ddfaf119

handle (processor_class, None) returned by ModelPatterns (#31753)

molbap committed 2 years ago

Verified c1fe1259

Adds final answer tool for all agents (#31703)

aymeric-roucher committed 2 years ago

Verified 0fd885b9

Requires for torch.tensor before casting (#31755)

echarlaix committed 2 years ago

Verified dc72fd7e

fix assisted decoding (#31401)

jiqing-feng committed 2 years ago

Verified 7f91f168

Fix documentation for Gemma2. (#31682)

jbornschein committed 2 years ago

Verified f91c16d2

Make tool JSON schemas consistent (#31756)

Rocketknight1 committed 2 years ago

Verified cd0935dd

🚨🚨 TextGenerationPipeline: rely on the tokenizer default kwargs (#31747)

gante committed 2 years ago

Verified 82486e59

Older