huggingface/transformers

Pull Requests Commits

Fix TIPSv2 text pooling dtype in bfloat16

tarekziade committed 3 days ago

04b0b58b

🚨 [serve][breaking] update attn for mac users (#49730)

SunMarc committed 4 days ago

Verified 66f51e76

Speed up `import transformers` by building the package distribution mapping lazily (#50047)

qgallouedec committed 4 days ago

Verified aba0d6e4

Restore the `MegaBlocksMoeMLP` hub kernel mapping for MXFP4 GPT-OSS (#50150)

jiqing-feng committed 4 days ago

Verified 955b6d35

Fix ZeRO-3 init for composite models built via _from_config (#49982)

kaixuanliu committed 4 days ago

Verified 5a5e9ef1

[docs] Quantization (#50093)

stevhliu committed 4 days ago

Verified f6d78a34

expose private parameter in `trainer.push_to_hub` (#49874)

not-lain committed 4 days ago

Verified 4293f191

Trainer: `save_pretrained` on every rank for PEFT model (#49316)

michaelbenayoun committed 4 days ago

Verified 17558143

[docs] Vision backend differences (#50131)

stevhliu committed 4 days ago

Verified 7501117d

fix(glm4v): bump backend_equivalence_atol to 0.5 for Pillow 13 compatibility (#50170)

ydshieh committed 4 days ago

Verified 27ae491e

Allow untokenized output from `apply_transcription_request` (#49944)

hmellor committed 4 days ago

Verified cc414c4f

need to check model_type and model_name (#50128)

itazap committed 4 days ago

Verified c5c5c55b

Add `TRANSFORMERS_SDPA_FORCE_GQA` envvar in `sdpa_attention_forward` (#49459)

jakobknauer committed 4 days ago

Verified c86af602

Fix `WatermarkDetector` p-value computation (#50123)

jonasschaefer committed 4 days ago

Verified 60893329

test: lazy-load pipeline classes to eliminate collection-phase noise (#50158)

ydshieh committed 4 days ago

Verified 4fed66c2

Remove torch.jit.script (#50082)

Rocketknight1 committed 4 days ago

Verified ce26dc87

Docs: build English only, other languages are machine-translated (#50120)

mishig25 committed 4 days ago

Verified 0a18165a

Mark generated files as linguist-generated (#50061)

ArthurZucker committed 4 days ago

Verified 506c3af7

Standardise processing requires (#50042)

guarin committed 4 days ago

Verified ccf2403e

Fix AttributeError on LinearAttentionLayer in export generate test (#50140)

ydshieh committed 4 days ago

Verified 0460c6d9

[Kosmos2] Fix SDPA applying a causal mask in the image-to-text projection (#50118)

tarekziade committed 4 days ago

Verified ad6950c9

v5.23.0-dev0 (#50155)

ArthurZucker committed 4 days ago

Verified d284899a

[LwDetr] Initialize encoder heads like the detection heads (#50041)

tarekziade committed 4 days ago

Verified 9e602e78

[OneFormer] Fix loss computation when is_training=False (#50039)

tarekziade committed 4 days ago

Verified 1cc4c969

Config annotations (#50067)

zucchini-nlp committed 4 days ago

Verified f90e93c4

Shard the parameters kept in float32 in their own FSDP2 group (#49501)

qgallouedec committed 4 days ago

Verified e82b00dd

Reuse the TP/EP test ranks across tests instead of spawning them per test (#49986)

3outeille committed 4 days ago

Verified ca47a0e1

🚨 build and expand once the tp/ep_plan/fsdp_plan (#49527)

3outeille committed 4 days ago

Verified 88e83834

handle local sink in TP (#49567)

3outeille committed 4 days ago

Verified 0a83ff7f

[Qwen4 Exp] Fix _no_split_modules naming a class that does not exist (#50138)

eilamc14 committed 4 days ago

Verified 4db5b946

Older