huggingface/transformers

Pull Requests Commits

tarekziade committed 39 minutes ago

cc6b2e28

Add experts_offload: stream MoE experts from disk through a bounded device cache

tarekziade committed 50 minutes ago

412192f6

CI pass, CI no pass? (#50690)

zucchini-nlp committed 3 hours ago

Verified 41f9d399

Add inkling2 (#50682)

guarin committed 15 hours ago

Verified 109a7d19

[`Gemma4`] Fix FA slicing (#50681)

vasqu committed 16 hours ago

Verified dc3bd32b

Restore the root-level PEFT adapter when resuming from a checkpoint w… (#50368)

ziywang50 committed 19 hours ago

Verified acf8004e

Granite for Docling (#49012)

nassarofficial committed 19 hours ago

Verified 46b887d9

🚨 Remove the `ep_router` expert-parallel style (#49615)

IlyasMoutawwakil committed 20 hours ago

Verified 506ac612

🚨 Run MXFP4 GPT-OSS experts through `ep_dispatch_experts` (#49613)

IlyasMoutawwakil committed 20 hours ago

Verified 95e9b5f6

Add dots3-note Preview model support (#47844)

miraclezqc committed 20 hours ago

Verified b0e6e7fd

Add IQuestCoder (#50579)

hmellor committed 20 hours ago

Verified bb98e254

Move the remaining router-masked MoE plans to `ep_dispatch_experts` (#49607)

IlyasMoutawwakil committed 21 hours ago

Verified 290c247f

🚨 `deepgemm_megamoe` EP implementation for DeepGEMM Mega MoE (#49606)

IlyasMoutawwakil committed 21 hours ago

Verified f5d8e7df

Add InternLM3 (#50578)

hmellor committed 21 hours ago

Verified 1c76d04b

fix(tokenizers): stop treating a missing transformers_version as Mistral (#49733)

cdcailf committed 22 hours ago

Verified 9b6e5e20

Load the released A.X-K1 checkpoints without unexpected or missing keys (#50644)

hmellor committed 22 hours ago

Verified 0923f62c

Fix get_json_schema truncating multi-line argument descriptions (#50606)

YusefSyed committed 22 hours ago

Verified e2427469

Make distributed Trainer tests actually resume from checkpoints (#49723)

Titus-von-Koeller committed 22 hours ago

Verified 27d7e6d2

Fix Mask2Former _init_weights to match Meta's original initialization (#50596)

KMaruthi2002 committed 23 hours ago

Verified 158f872e

Fix Inkling router precision in BF16 and autocast (#50621)

taking-lying-flat committed 23 hours ago

Verified 52030880

QA: Make `make run-mic` easier to read and to target (#50652)

tarekziade committed 1 day ago

Verified 40a1fb90

Fix `HfArgumentParser` on Python 3.15 (#50651)

qgallouedec committed 1 day ago

Verified 3d284ce0

Avoid mutating input examples in DataCollatorForMultipleChoice (#50611)

LKL-ZREO committed 1 day ago

Verified 5c239eec

Fix XVector outputs of the wav2vec2 family when the TDNN layers are offloaded (#50646)

ydshieh committed 1 day ago

Verified fde3a351

[`Offloading`] Fix tests for some models (#50631)

vasqu committed 1 day ago

Verified 06edaf11

[`FA`] Fix FA tests for unique models (#50632)

vasqu committed 1 day ago

Verified 1ee78b6e

🚨 Slim down `[audio]` to torchaudio + soundfile + soxr (drop librosa and phonemizer) (#50610)

eustlb committed 1 day ago

Verified 880561aa

[`GptOss`] Fix the kernels autodispatch test (#50630)

vasqu committed 1 day ago

Verified f61791ca

Expert-parallel implementations for `ep_dispatch_experts`: all-reduce at inference, all-to-all in training (#49429)

IlyasMoutawwakil committed 1 day ago

Verified b7394bc7

Exporters: run exported models on their backend's runtime (`ModelRunner`) (#50032)

IlyasMoutawwakil committed 1 day ago

Verified 57583c28

Older