huggingface/transformers

Pull Requests Commits

[CI] AMD docker test: build kenlm from source with regenerated Cython (py3.14)

Abdennacer-Badaoui committed 26 days ago

b6bd85a9

[CI] AMD docker test: use CPU torchcodec build (PyPI wheel needs CUDA libnvrtc) + import check

Abdennacer-Badaoui committed 27 days ago

b7eba427

[CI] AMD docker test: push to huggingface/transformers-pytorch-amd-gpu:rocm10-test only

Abdennacer-Badaoui committed 27 days ago

360a90b1

[CI] AMD docker test: drop GitHub API cache-bust ADD (403 rate limit on runner)

Abdennacer-Badaoui committed 27 days ago

357d3387

[CI] AMD docker test: rocm/vllm rocm10.0.0 / py3.14 / torch 2.12 base (build only, no push)

Abdennacer-Badaoui committed 27 days ago

2d8ad9eb

[Fix] Remove old test file with two ancient tests (#49280)

remi-or committed 27 days ago

Verified f5ab8561

Fix chunked prefill with multi-axis position ids (Qwen3-VL) (#49319)

dacorvo committed 27 days ago

Verified 2c8526da

Fix multi-image feature ordering during generation batch expansion (#49276)

DimensionSTP committed 28 days ago

Verified 9d68278d

Fix quantized cache (#48700)

jiqing-feng committed 28 days ago

Verified 492f2404

Why override when you can fix in `Mixin` (#49105)

zucchini-nlp committed 28 days ago

Verified 3d207f81

Do RoPE with mul instead of matmul (#49265)

Rocketknight1 committed 28 days ago

Verified c633787a

Deprecated pipelines should redirect; removed pipelines should raise (#49308)

LysandreJik committed 28 days ago

Verified 56d3afc0

Enable model test on upcoming `torch_tpu` backend (#49207)

tengomucho committed 28 days ago

Verified 26c0a7fd

Fix encoder repetition penalty source rows after batch expansion (#49262)

gitedmond committed 28 days ago

Verified 10502571

Preserve masked EOS scores in exponential decay length penalty (#49260)

gitedmond committed 28 days ago

Verified 1aac9583

[CI] Avoid fetching all artifacts when only specific ones are needed (#49282)

ydshieh committed 29 days ago

Verified 46923035

[CI] check_bad_commit: use EFS cache to avoid Xet FUSE OOM (exit 137) (#49273)

ydshieh committed 30 days ago

Verified 02d8fb97

🚨 Select the OWLv2 image query by objectness (#49200)

qgallouedec committed 30 days ago

Verified 42adb294

Revert "[CI] ssh-runner: add optional cache_type input to switch between bucket and EFS runners (#49159)" (#49270)

ydshieh committed 30 days ago

Verified 3693f8d2

Add Trainer.loss_is_scaled_for_ga to declare whether compute_loss already scales for gradient accumulation (#49240)

qgallouedec committed 30 days ago

Verified 094f7101

Don't forward text-level flash-attention kwargs to multimodal encoders (#49227)

Abdennacer-Badaoui committed 31 days ago

Verified 35dff095

Fix pp chart2table inference (#49225)

zucchini-nlp committed 31 days ago

Verified 9d8c3a24

fix precomputed flash kwargs clashing (#47928)

IlyasMoutawwakil committed 31 days ago

Verified 000a29ff

Support per-layer cache configuration (#48178)

eladsegal committed 31 days ago

Verified 35924ec3

Reapply modular examples (#49230)

Cyrilvallez committed 31 days ago

Verified 99168b2a

Remove some stuff that does not exist (#49232)

Cyrilvallez committed 31 days ago

Verified fe4107c3

Remove deprecation cycle on mask and cache (#49231)

Cyrilvallez committed 31 days ago

Verified f9ecae3e

Remove rotaries deprecation cycle (#49228)

Cyrilvallez committed 31 days ago

Verified 0a896aa4

[`GTE`] Add memory mixin (#49234)

vasqu committed 31 days ago

Verified a005fc82

[CB] Make CB more device agnostic and add support for XPU (#49156)

remi-or committed 31 days ago

Verified 0bb187fe

Older