huggingface/transformers

Pull Requests Commits

Merge branch 'main' into ep-token-dispatch

3outeille committed 60 days ago

Verified 56047f4c

[docs] mrope and axial rope (#48717)

stevhliu committed 60 days ago

Verified 2cba1950

Fix `AutoModel.from_pretrained` not restoring `modules_to_save` weights (#48595)

shniubobo committed 60 days ago

Verified dbbd5512

Add integration tests for MuseGlimmerAssistantModel (#48796)

ydshieh committed 60 days ago

Verified 39f81b04

Fix kosmos flaky test (#48800)

IlyasMoutawwakil committed 60 days ago

Verified 229cd066

Restore legacy tensor-parallel initialization for compatibility (#48797)

3outeille committed 60 days ago

Verified a1ba1c14

[serge] Fix 12 integration tests for model `edgetam` failing with `import_or_config` (other (12)) (#48322)

sergereview[bot] committed 60 days ago

Verified d72aef09

Bump `peft` version requirement (#48716)

shniubobo committed 60 days ago

Verified 75c3583e

add kernel mapping entry for RMSNormGated, KDA, Conv1D on XPU (#48702)

kaixuanliu committed 60 days ago

Verified ca3bdf7a

Cast pixel values to the patch embedding dtype in DeepSeek-OCR-2 (#48632)

jiqing-feng committed 60 days ago

Verified a20852a0

Fix TextToAudioPipeline crash for tokenizer-only models (#48505)

jiqing-feng committed 60 days ago

Verified 9eb65c32

Fix inkling embedding norm (#48786)

Cyrilvallez committed 60 days ago

Verified 33849085

Fix hidden state selection on Gemma4 assistant's first prefill step (#48704)

glistening committed 60 days ago

Verified ff377ea4

Kernel api doc, part 2 (#46889)

michaelbenayoun committed 60 days ago

Verified 7219d282

[InternVL] Normalize num_patches before np.cumsum (#48469)

lorenzozanee committed 60 days ago

Verified 9ab9957d

Fix Glm4vMoeIntegrationTest: offload_folder + MemoryCleanupMixin (#48776)

ydshieh committed 60 days ago

Verified f24b457b

Fix silently random-initializing `RTDetrModel`/`SEWDForCTC` loads (wrong `base_model_prefix`) (#48744)

abhinav-phi committed 60 days ago

Verified 56c5e876

fix videomae load error (#48675)

sywangyi committed 60 days ago

Verified 06a7a68e

Switch daily CI to torch 2.14 — update expected outputs (#48750)

ydshieh committed 61 days ago

Verified a2c15b30

Give the token dispatch weight comparison its own tolerance

qgallouedec committed 61 days ago

86561b69

Keep reading the tensor parallel size from accelerate in `get_tp_size`

qgallouedec committed 61 days ago

4f2a6b91

Address review: guard dispatch + PP, derive the mesh name, move the guards

qgallouedec committed 61 days ago

3439765f

Merge remote-tracking branch 'upstream/main' into ep-token-dispatch

qgallouedec committed 61 days ago

b169a75d

Merge remote-tracking branch 'upstream/fsdp2-ep-2d-mesh' into ep-token-dispatch

qgallouedec committed 62 days ago

d604a2a2

Enable FSDP2 + expert parallelism via a 2-D (fsdp, tp) device mesh (#48516)

qgallouedec committed 62 days ago

Verified 415e6d2f

Trim the parallelism config comment

qgallouedec committed 62 days ago

0147840f

Drop the comment on the parallel sizes

qgallouedec committed 62 days ago

51207488

Merge remote-tracking branch 'upstream/main' into fsdp2-ep-2d-mesh

qgallouedec committed 62 days ago

79988c35

unifying device_mesh init to enable PP + TP inference (#48155)

3outeille committed 62 days ago

Verified 5474a55e

Use huggingface_hub httpx export (#48685)

Wauplin committed 62 days ago

Verified 80a79d1b

Older