huggingface/transformers

Pull Requests Commits

fix TP issue within the device mesh (Tensor Parallel group) (#43105)

sywangyi committed 89 days ago

Verified 314a4515

Fix Qwen3OmniMoe Talker weight loading and config initialization (#43084)

HuiyingLi committed 89 days ago

Verified 64a476b7

Improve `_init_weights` tests (#43108)

Cyrilvallez committed 89 days ago

Verified 486a7229

Fix Qwen3-VL typos and clarify image/video/vision token descriptions (#43033)

stefgina committed 89 days ago

Verified 37c9c941

fix(qwen2vl): make Qwen2VLImageProcessor respect config['size'], align with fast version (#43095)

qianyue76 committed 89 days ago

Verified b19834d8

batched and grouped experts implementations (#42697)

IlyasMoutawwakil committed 89 days ago

Verified 0642963b

Make projection LayerNorm configurable in LFM2-VL (#43087)

ankke committed 90 days ago

Verified 3c251772

GLM-ASR Support (#42875)

zRzRzRzRzRzRzR committed 101 days ago

Verified a7f29523

Fix audio pipelines (#42920)

zucchini-nlp committed 102 days ago

Verified 7bbfc19a

Move missing weights and non-persistent buffers to correct device earlier (#43021)

Cyrilvallez committed 102 days ago

Verified d6a6c826

Do not use accelerate hooks if the device_map has only 1 device (#43019)

Cyrilvallez committed 102 days ago

Verified 315dcbe4

Ensure CI runs `check_modeling_structure` and fixes for `make fixup` (#43002)

ebezzam committed 102 days ago

Verified dc06f2dd

Torch distrib smallfix (#43008)

molbap committed 102 days ago

Verified 50ddcfd2

Prefill-related logic in input preparation for generation (#42088)

zucchini-nlp committed 103 days ago

Verified ccc7d90b

Move `original_max_position_embeddings` to rope params (#42513)

zucchini-nlp committed 103 days ago

Verified 34732810

Do not return a tuple in mistral tokenizer Automapping (#42997)

molbap committed 103 days ago

Verified 9971e410

Fix deepspeed + quantization (#43006)

SunMarc committed 103 days ago

Verified 42512f79

[loading][TP] Fix device placement at loading-time, and simplify sharding primitives (#43003)

Cyrilvallez committed 103 days ago

Verified 5f1c05cf

Improve spacing of markdown files (#42984)

cyyever committed 103 days ago

Verified e9f0f8e0

Fix incorrect library name in BitNet integration warning (#42966)

leaderofARS committed 103 days ago

Verified f7d139c1

Fix DocQA max_answer_len validation error message (#42948)

antznette1 committed 103 days ago

Verified 64ec2ba1

Fix formatting of trackio model tag (#42973)

abidlabs committed 103 days ago

Verified 817886a6

Hardcode the factor in caching allocator (#42996)

Cyrilvallez committed 103 days ago

Verified f5d9d808

Do not use global variable, and improve context manager coverage (#42957)

Cyrilvallez committed 103 days ago

Verified 0d2dbaa9

[`Tests`] Fix inputs placement (#42963)

vasqu committed 104 days ago

Verified 007274db

fix device dismatch issue for pe_audio_video model parallelism (#42917)

kaixuanliu committed 104 days ago

Verified 1b280f99

Single config attribute for weight tying (#42815)

zucchini-nlp committed 104 days ago

Verified 9a90500b

Fix tests trainer again (#42933)

SunMarc committed 106 days ago

Verified 70179949

[`Ernie 4.5`] Ernie VL models (#39585)

vasqu committed 106 days ago

Verified a8a22624

[`Generation`] Fix default overwrite for non-`None` defaults (#42958)

vasqu committed 106 days ago

Verified f218ed21

Newer Older