huggingface/transformers

Pull Requests Commits

Run fine-grained FP8 expert parallelism as masked expert parallelism

IlyasMoutawwakil committed 1 day ago

63a715c0

skip `sdpa_can_dispatch_on_flash` tests for 3 models (#49809)

kaixuanliu committed 1 day ago

Verified acfee829

🚨Add an FP8 backend to `QuantizedCache` (#48677)

jiqing-feng committed 1 day ago

Verified 43cc0abd

[LED][Longformer] Replace for-loop with unfold in _chunk ONNX-export path (#46169)

guinik committed 2 days ago

Verified 17c6736c

logits_to_keep bool mask for all ForXXX classes (#48825)

jp1924 committed 2 days ago

Verified a672eb5f

[Fetcher] Fix missing test deps for dynamically-resolved image processors (#49826)

ydshieh committed 2 days ago

Verified bf6c200f

Fix several object detection models/loader fix for module tying (#49827)

molbap committed 2 days ago

Verified bba5c055

Fix encoder logits processors when precomputed encoder_outputs lack token IDs (#49696)

kunalsanga committed 2 days ago

Verified fb3e4f28

Defer some imports (#49663)

jayzuccarelli committed 2 days ago

Verified 30a8894c

refactor CTC loss (#49833)

ebezzam committed 2 days ago

Verified ab7e5ac3

Add Intel XPU CI (#49254)

regisss committed 2 days ago

Verified 68f60eac

fix(mask2former): register input_projections as nn.ModuleList (fixes #49698) (#49703)

Omarinho12345 committed 2 days ago

Verified f348d3dc

Add tensor parallel plan for Nemotron (#48780)

Siddhant231xyz committed 2 days ago

Verified 8e40ad7b

Only tie MM Grounding DINO decoder heads when the config shares them (#49552)

lu123bing committed 2 days ago

Verified d63debe3

Fix Windows path separators breaking relative imports of Hub custom-generate modules (#48951)

dltsum committed 2 days ago

Verified 857d1a0c

Add UltraVBERT (#49685)

athrael-soju committed 2 days ago

Verified 8563d78e

fix(vision): Route the DETR conversion scripts through from_pretrained (#49830)

harshaljanjani committed 2 days ago

Verified 6e280fbf

[Docs] Add Marathi (मराठी) translation of README (#48746)

shreeyadav06 committed 2 days ago

Verified 6a994bc4

Fix multi-zoom crops in the SAM mask-generation pipeline and related bugs (Fixes #35530) (#49689)

stlukyanenko committed 2 days ago

Verified 0b5cfc26

fix(audio): Route the ALM conversion scripts through from_pretrained (#48525)

harshaljanjani committed 2 days ago

Verified 72021435

fix model parallel bug for molmo2 model (#49816)

kaixuanliu committed 2 days ago

Verified 481e371f

Add Flint model support. (#49824)

lovishmadaan committed 2 days ago

Verified 0ea108da

Remove sdpa support for models using softcapping (#49811)

qgallouedec committed 2 days ago

Verified 7d3ae42e

Fix CLVP `c_proj` residual scaled-init silently never applied (#47478)

ShamSaleem committed 2 days ago

Verified 5fd67cf1

fix failed test cases for unlimited_ocr model (#49806)

kaixuanliu committed 2 days ago

Verified 82307147

Fix dtype cast in TimesFM 2.5 ResidualBlock for quantized weights (#46942)

JOhnsonKC201 committed 2 days ago

Verified 0d65c9cd

Fix transformers env when requests is not installed (#48412)

panshularora committed 2 days ago

Verified e2fec23a

Fix `base_model_prefix` in the PP-OCR detectors, CLIPSeg, TimesFM, CLIP and VideoPrism heads so base checkpoints load (#48871)

blipbyte committed 2 days ago

Verified c2abdef3

Apply EXIF orientation in load_image_as_tensor with torchvision (#49794)

digit50 committed 2 days ago

Verified d38514c6

Fix Fuyu processor test (#49821)

zucchini-nlp committed 2 days ago

Verified a762035c

Older