Commits vllm-project/vllm

optimize allpool forward

yewentao256 committed 108 days ago

b870c8ed

[Docs] Add non-causal support to attention backend docs (#41643)

MatthewBonanni committed 108 days ago

Verified be5983b8

[NVFP4][fix] Fix `layer.weight` -> `w13` typo in NVFP4 MOE emulation kernel preparation (#41630)

fxmarty-amd committed 108 days ago

Verified 9c07342f

feat: update xgrammar==0.2.0 to use structural tags for strict tool calling + reasoning for more models (#40894)

Seven-Streams committed 108 days ago

Verified 844df542

[bugfix] Fix prompt logprobs on request eviction during chunked prefill (#41411)

joa-stdn committed 108 days ago

Verified 422dd025

Fix Nano Nemotron text-only weight loading (#41205)

Baekpica committed 108 days ago

Verified 8c780943

[ROCm] ROCm7.2.2 + profiler fix + AITER 0.1.12.post2 (#41386)

gshtras committed 108 days ago

Verified e724b0ea

[Bugfix] KimiK2ReasoningParser: guard against buffered end-token in streaming (#41068)

JasonKeyiL committed 108 days ago

Verified 712ad028

Limit gpu utils and lower max BS on test_transcription_api_correctness.py (#41649)

ekagra-ranjan committed 108 days ago

Verified 321fa2d6

[Bug] Fix `tests/compile/test_config.py` AttributeError: 'NoneType' object has no attribute 'dtype' (#41288)

yewentao256 committed 108 days ago

Verified 3e1ad443

Test nemotron nano-v2 and nemotron nano-v3 separately, disable super-omni redundant tests (#41616)

netanel-haber committed 108 days ago

Verified 8decbfa0

Revert "[Doc] Fix RTD build: pytorch.org/docs/stable/objects.inv returns 404" (#41618)

stecasta committed 108 days ago

Verified 62ba7516

[Bugfix] Apply ruff-format to hyperclovax.py (#41620)

stecasta committed 108 days ago

Verified 6f53753f

[CI] Stabilize cpu offload compressed tensors test (#41102)

AndreasKaratzas committed 109 days ago

Verified 6ec9bbec

[ROCm][CI] Align spec decode logprob test prefill settings (#41335)

AndreasKaratzas committed 109 days ago

Verified 01d4d1ad

[Transformers v5] Vendor HCXVisionConfig for compatibility (#38447)

HanFa committed 109 days ago

Verified c103c02a

[CI] Clean up remote servers on pytest parent exit (#41570)

AndreasKaratzas committed 109 days ago

Verified 67058ca3

[Bench] Forward --seed to CustomDataset and CustomMMDataset shuffle (#40788)

Aktsvigun committed 109 days ago

Verified 894a0250

[Bugfix] Fix degenerate KV cache stride causing TMA cudaErrorIllegalInstruction (#40737)

the-david-oy committed 109 days ago

Verified 66dfee71

[Bugfix] Fix FP8 Bias Loading (#41424)

alex-jw-brooks committed 109 days ago

Verified db9a84e0

[Bugfix][Ray] Fix RayExecutorV2 actor name collision with DP > 1 (#40398)

tomeras91 committed 109 days ago

Verified cb03fee3

Disable flashinfer autotune temporarily due to correctness issues (#41524)

wzhao18 committed 109 days ago

Verified c51df430

[Doc] Add Qwen3-30B-A3B-Thinking-2507-FP8 to batch invariance verified models (#41513)

taneem-ibrahim committed 109 days ago

Verified 54dc64d5

[MRV2] Add shutdown() method (#41297)

WoosukKwon committed 110 days ago

Verified e6ff3e9c

[Quantization] add humming mxfp4 moe backend (#41083)

jinzhen-lin committed 110 days ago

Verified 08834cc3

[DSv4] Tune default value of `VLLM_MULTI_STREAM_GEMM_TOKEN_THRESHOLD` (#41526)

ywang96 committed 110 days ago

Verified 856ec480

[DSV4] Guard megamoe flag with Pure TP (#41522)

zyongye committed 110 days ago

Verified 1c607d7b

[CI] Add ci-fetch-log.sh helper for Buildkite job logs (#41517)

mgoin committed 110 days ago

Verified 4f7309fc

Revert "[Build] Make bundled DeepGEMM wheel portable across Python versions" (#41512)

mgoin committed 110 days ago

Verified 0a9362d6

[Build] Switch CUDA 13.0 wheel builds to PyTorch manylinux_2_28 base (#41416)

mgoin committed 110 days ago

Verified cfd2573f