huggingface/transformers

Pull Requests Commits

Port Llama 4's MoE to the shared experts interface

IlyasMoutawwakil committed 1 day ago

d607e06b

Fix Whisper batch timestamp decoding for tensors (#49419)

Kingofbooks committed 1 day ago

Verified 9167f73f

:rotating_light: SubConfigSpec for centralized init of subconfigs (#48282)

zucchini-nlp committed 1 day ago

Verified d4f69a86

PR CI: grant actions: read to the reusable workflow (#49417)

tarekziade committed 2 days ago

Verified 125f265a

[Bugfix] [Modeling] Fix DSV4 norm (#47486)

kylesayrs committed 2 days ago

Verified fa406191

Do not wrap the model in FSDP for expert-parallel token dispatch (#49383)

3outeille committed 2 days ago

Verified cb33194a

[docs] Update Spaces (#49242)

stevhliu committed 2 days ago

Verified a96730c8

handle empty precompiled charsmap in spm converter (#49211)

itazap committed 2 days ago

Verified 91ea1192

🚨🚨🚨 luke fixe (#48408)

itazap committed 2 days ago

Verified 2b727b6d

[CI] Temporarily disable AMD scheduled CI trigger (#49394)

ydshieh committed 2 days ago

Verified 7b822d20

[tests] Fix flaky CMVN tests in Speech2TextFeatureExtractionWithoutTorchaudioTest (#49374)

ydshieh committed 2 days ago

Verified 705a7cf8

Fix Phi3 LongRoPE cache rebuild at the context boundary (#49361)

LE0-Lin committed 2 days ago

Verified 6a33cb58

[Docker] Upgrade to torch 2.14/CUDA 13.0 and optimize FA2 build from source (#49381)

ydshieh committed 2 days ago

Verified 6d3802a4

Fix Pixtral Flash Attention image boundaries after axial RoPE refactor (#49373)

oliverholworthy committed 2 days ago

Verified ff704ffd

Let token dispatch run with `ep_size != tp_size` (#49368)

qgallouedec committed 3 days ago

Verified e598fbad

[tests] Skip onnx.exactness for BitForImageClassification (ORT fp32 drift) (#49366)

ydshieh committed 3 days ago

Verified 14e738b5

Link the padding side section from SSM and hybrid model docs (#49247)

jayzuccarelli committed 3 days ago

Verified cbde22f4

Smthn smthn (#49364)

vasqu committed 3 days ago

Verified 92cd495f

Revert "[CI] Temporarily disable AMD scheduled CI trigger" (#49271) (#49362)

ydshieh committed 3 days ago

Verified a7e8ff2b

ALM testx fIxing... (#49355)

zucchini-nlp committed 3 days ago

Verified 571c5cc7

Fix DataCollatorForLanguageModeling ignoring seed=0 (#49349)

akanyaani committed 3 days ago

Verified b4526a76

Fix explicit kernelization modes for evaluation models (#49339)

DimensionSTP committed 3 days ago

Verified 5a27d2d2

fix(cohere_compass): add video to input_modalities (#49320) (#49323)

destopianpirate committed 3 days ago

Verified 23c21ac8

GPT-OSS: read the clamped SwiGLU's alpha and limit from the config (#49346)

IlyasMoutawwakil committed 3 days ago

Verified f1d0d9fb

[`distributed`] Default MoE `ep_plans` to token dispatch (#49160)

3outeille committed 3 days ago

Verified 9921402b

[`distributed`]: adapt Trainer to work Expert Parallel (#48873)

3outeille committed 3 days ago

Verified 8625785e

Move CI to Python 3.11, with the version set in `.python-version` (#49311)

tarekziade committed 4 days ago

Verified 422bd07e

Bump doc-builder main docs workflow pin (#49350)

paulinebm committed 4 days ago

Verified 03011bd6

[`distributed`] Add expert-parallel token dispatch, default for Qwen3 MoE (#48865)

3outeille committed 4 days ago

Verified d28b7937

[`distributed`] decoupled tp_plan and ep_plan (#48859)

3outeille committed 4 days ago

Verified 2033cf02

Older