huggingface/transformers

Pull Requests Commits

Prefetch the shards in the background instead of waiting for them

qgallouedec committed 3 days ago

Verified d87e79cc

Bind the prefetch barrier to the accelerator

qgallouedec committed 3 days ago

Verified cd6ad0ec

Apply ruff format to the prefetch helper

qgallouedec committed 3 days ago

Verified d17e1335

Move the shard prefetch into distributed/utils.py

qgallouedec committed 3 days ago

Verified f04f7a08

qgallouedec committed 3 days ago

Verified 05ca7a73

Warm the page cache before sharded loading (opt-in via HF_SHARD_PREFETCH)

qgallouedec committed 3 days ago

Verified a07648d4

Rename `gradient_checkpointing` to `activation_checkpointing` in the public API (#49248)

qgallouedec committed 3 days ago

Verified a0cf918f

[`Moe`] Fix grouped mm fallback (#49857)

vasqu committed 3 days ago

Verified c6fab466

🚨 Safe-ize pi0 graphs (#50160)

molbap committed 3 days ago

Verified 865e28f9

Chat template output tensor type (#50103)

zucchini-nlp committed 4 days ago

Verified 49368678

[Distributed] Set `device_id` only for NCCL (#50281)

michaelbenayoun committed 4 days ago

Verified 409c3490

[Trainer] Compile with static shapes when the backend is neuron (#49617)

qgallouedec committed 4 days ago

Verified 26754696

[Improvement] Formalize KV sharing (#47290)

remi-or committed 4 days ago

Verified cc1451df

Docs: remove duplicate hf_token secret in build_documentation.yml (#50168)

mishig25 committed 4 days ago

Verified 4888db4e

Fix `bboxes` being overwritten before use in `prepare_example` (#50218)

Jose-Marcial-GF committed 4 days ago

Verified c97587a4

fix(tests_fetcher): handle `X as X` re-export syntax and duplicate module entries (#50253)

ydshieh committed 4 days ago

Verified e03c3539

Read add_bos_token and add_eos_token from GGUF files (#49649)

francip committed 4 days ago

Verified 7c22a03f

[serve] Respect max_completion_tokens in /v1/chat/completions (#50225)

fruitstash committed 4 days ago

Verified a9124a87

Dequantize MXFP4 experts to the requested dtype (#50256)

qgallouedec committed 4 days ago

Verified 67685971

deal with Nan values for aggregating loss (#50172)

SunMarc committed 4 days ago

Verified b47ec788

Fix OOM regression for FSDP2 + cpu_ram_efficient_loading on large models (#45649)

AmineDiro committed 4 days ago

Verified 5949d2c4

[serge] Fix 2 integration tests regressed by commit 7d3ae42ee810 (PR #49811) (#50098)

sergereview[bot] committed 4 days ago

Verified 763b4ba5

Fix resize_embeddings tests for deepspeed (#49909)

kaixuanliu committed 4 days ago

Verified 28aa8646

[VibeVoice] Add `sample` argument to disable VAE noise in audio features (#50179)

hmellor committed 4 days ago

Verified 2f2e711f

fix failed test cases for pocket_tts model (#50159)

kaixuanliu committed 4 days ago

Verified 3e8d3971

fix(tests_fetcher): use posix paths throughout, eliminate all os.path.sep/os.path.join patterns (#50219)

ydshieh committed 4 days ago

Verified a8f6bda5

Remove the GitHub API cache-bust from the AMD docker images (#49351)

tarekziade committed 4 days ago

Verified 044aa4c8

[Gemma4] Pass num_frames_per_video to get_video_features in forward (#50237)

yzkkai committed 4 days ago

Verified 5f6629d8

fix Load DETR-family panoptic masks with load_image_as_tensor (#50234)

dhekimian committed 4 days ago

Verified 9167cfce

update to torch 2.15 for xpu (#50236)

sywangyi committed 4 days ago

Verified 7db0255f

Older