huggingface/transformers

Pull Requests Commits

Merge branch 'make-fa-test-less-flaky' into fix-fa-head-dims

vasqu committed 5 days ago

8b81802e

vasqu committed 5 days ago

bea23ca6

Merge branch 'fix-old-fa-models' into make-fa-test-less-flaky

vasqu committed 5 days ago

dbc0e9c7

this should not be a constant

vasqu committed 5 days ago

1ee77d79

vasqu committed 5 days ago

733a6ba5

Merge branch 'make-fa-test-less-flaky' into fix-fa-head-dims

vasqu committed 5 days ago

2d7abff1

Merge branch 'fix-old-fa-models' into make-fa-test-less-flaky

vasqu committed 5 days ago

f6323a13

Merge branch 'fix-fa-compile-cache' into fix-old-fa-models

vasqu committed 5 days ago

5d7bb376

Merge branch 'fix-moe-fallback-bwd-compile' into fix-fa-compile-cache

vasqu committed 5 days ago

a82f7c1c

vasqu committed 5 days ago

2fe7fe15

fix models head dim

vasqu committed 5 days ago

8742f5ac

vasqu committed 5 days ago

b00d4737

vasqu committed 5 days ago

232743af

fix fa static cache

vasqu committed 5 days ago

fc2b455e

fix fallback bwd

vasqu committed 5 days ago

8fdb4534

Fine-grained FP8: DeepSeek-V4's expert_dtype in the quantization config (#49848)

IlyasMoutawwakil committed 5 days ago

Verified 9a6925a9

Upgrade Python Min version to `3.11` (#49251)

Sai-Suraj-27 committed 5 days ago

Verified 79557e27

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

kaixuanliu committed 5 days ago

Verified acfee829

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

jiqing-feng committed 5 days ago

Verified 43cc0abd

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

guinik committed 5 days ago

Verified 17c6736c

logits_to_keep bool mask for all ForXXX classes (#48825)

jp1924 committed 5 days ago

Verified a672eb5f

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

ydshieh committed 5 days ago

Verified bf6c200f

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

molbap committed 5 days ago

Verified bba5c055

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

kunalsanga committed 5 days ago

Verified fb3e4f28

Defer some imports (#49663)

jayzuccarelli committed 5 days ago

Verified 30a8894c

refactor CTC loss (#49833)

ebezzam committed 5 days ago

Verified ab7e5ac3

Add Intel XPU CI (#49254)

regisss committed 5 days ago

Verified 68f60eac

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

Omarinho12345 committed 6 days ago

Verified f348d3dc

Add tensor parallel plan for Nemotron (#48780)

Siddhant231xyz committed 6 days ago

Verified 8e40ad7b

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

lu123bing committed 6 days ago

Verified d63debe3

Older