huggingface/transformers

Pull Requests Commits

test muse sparknvfp4 after merge

IlyasMoutawwakil committed 26 days ago

585946e8

Merge main into finegrained-per-scheme-demo

IlyasMoutawwakil committed 26 days ago

9620ef5a

Add Muse Spark Open (#49644)

guarin committed 26 days ago

Verified 487dcfa6

[docs] Hardware docs template (#49575)

stevhliu committed 26 days ago

Verified d2e102b6

Add MOSS-Transcribe-Diarize model (#48386)

itazap committed 26 days ago

Verified 9d59c2c5

Integrate the UltraBERT architecture (#49596)

tomaarsen committed 26 days ago

Verified 4c6f2559

Add MIC contract lock files for PR CI integration (#49631)

ydshieh committed 26 days ago

Verified b0eff172

FSDP x TP Saving / Loading, Optimizer (#49256)

michaelbenayoun committed 26 days ago

Verified e62ee803

Add Unlimited OCR (#46836)

guarin committed 26 days ago

Verified 5789b6dd

FSDP x TP Saving / Loading (#48802)

michaelbenayoun committed 26 days ago

Verified 89d1c880

Fine-grained tests: per-expert gate_up activation globals, and the operands helper on the mock

IlyasMoutawwakil committed 26 days ago

4e8aa7e7

Merge origin/main into finegrained-per-scheme-demo

IlyasMoutawwakil committed 26 days ago

71ee0441

Use aiter-rope version 3 on ROCm (#49595)

Abdennacer-Badaoui committed 26 days ago

Verified 2b8b231f

Don't set a device per rank for gloo tensor-parallel tests (#49636)

IlyasMoutawwakil committed 26 days ago

Verified 276d3fac

fix(distributed): keep enable_expert_parallel as EP alone (#49633)

dacorvo committed 26 days ago

Verified d21c0d9e

Remove missing methods from docs (#49637)

guarin committed 26 days ago

Verified 16cad282

Translate `torch.mm(..., out_dtype=...)` to ONNX (#49635)

IlyasMoutawwakil committed 26 days ago

Verified 0e85988b

[docs] Hardware section (#49541)

michaelbenayoun committed 27 days ago

Verified c9e5f15a

Raise when a torch fallback would drop packed-sequence boundaries (#49474)

qgallouedec committed 27 days ago

Verified 01f16c0e

Fa packed sequence decode (#49610)

SunMarc committed 27 days ago

Verified da23990a

fix nested set_attn_implementation (#49138)

eustlb committed 27 days ago

Verified 4e3fc5e8

feat: Add MTP support to Mellum2/2.1 (#49531)

shadeMe committed 27 days ago

Verified 14300401

Fall back to the AITER kernel for flash_attention_2 on ROCm (#49608)

Abdennacer-Badaoui committed 27 days ago

Verified 2126f65b

extend onnx exporters tests to non-cuda devices (#49627)

kaixuanliu committed 27 days ago

Verified cd1d4736

Document video processor backends (#49470)

DimensionSTP committed 27 days ago

Verified b4c24288

Make sinusoidal position embeddings compileable (#49482)

jiqing-feng committed 27 days ago

Verified 24a2afba

Fix a port allocation race in FSDP tests (#49455)

icavan committed 27 days ago

Verified e24eb546

Save the PEFT adapter of a load-time sharded model from its own parameters (#49500)

qgallouedec committed 27 days ago

Verified f0705f99

fix: Pass the MTP layer type to per-layer-type rotary embeddings (#49468)

shadeMe committed 27 days ago

Verified 637b76e4

🚨 Remove deprecated cache utils methods (#49605)

guarin committed 27 days ago

Verified 59a50426

Older