vllm-project/vllm

Pull Requests Commits

[Frontend] Consolidate scale out entrypoints (#44512)

noooop committed 5 days ago

Verified 3483240b

[Perf][2/N] Expand Triton kernel warmup coverage, Qwen (#46750)

LopezCastroRoberto committed 5 days ago

Verified eddfd4cf

[mypy] Enable mypy for tests directory (#47018)

hickeyma committed 5 days ago

Verified a4e3cb40

Fix model info cache for package models (#46567)

soaringk committed 5 days ago

Verified ab132ee9

[Bugfix] Use native SiLU activation in CPU fused MoE (#45961)

aldenlobo committed 5 days ago

Verified e1861078

[Bugfix] Transformers backend: apply learned lm_head.bias for tied-embedding models (#46835)

JohnLangford committed 5 days ago

Verified 0e207dac

[CI Failure] Add transformers version check for openai/privacy-filter (#47011)

noooop committed 5 days ago

Verified 9e86352c

Remove unnecessary `load_weights` methods (#44589)

hmellor committed 5 days ago

Verified 5051698e

[ROCm][CI] Explicitly tear down multimodal offline LLMs (#46999)

AndreasKaratzas committed 5 days ago

Verified db28ae2d

Fix docs on main (#47009)

hmellor committed 5 days ago

Verified f6bb8682

[MM][CG] Gemma3 Encoder CUDA Graph (#43591)

JisoLya committed 5 days ago

Verified 4559c43a

[Rust Frontend] Add Harmony Renderer for GPT-OSS (#46800)

BugenZhao committed 5 days ago

Verified 5274c118

[CPU] Support cpu compressed-tensor w8a8 int8 moe (#42920)

yuwenzho committed 5 days ago

Verified 58d6a6e6

[EPLB] Mask padding in EPLB load recording (#38128)

ilmarkov committed 5 days ago

Verified a2abce64

Remove boilerplate missed by #46820 (#46956)

hmellor committed 5 days ago

Verified 311ad689

[Spec Decode] Avoid redundant hidden-states gather in draft prefill (#46968)

WoosukKwon committed 5 days ago

Verified 04724365

[Model] Add support for openai/privacy-filter (#41026)

fjosw committed 5 days ago

Verified 4dfbf150

[Bugfix][Mooncake] Fix Mooncake lookup prefixes with DCP > 1 (#46855)

wzhao18 committed 5 days ago

Verified 95528527

[ROCm][CI] Fix `rlhf_async_new_apis` Example On ROCm (#46895)

micah-wil committed 5 days ago

Verified c2127a25

[OCP MX ] Add back emulation to available OCP MX backends list (#46629)

fxmarty-amd committed 5 days ago

Verified 03c6d01c

[GLM5] Fix minor typo (#46961)

WoosukKwon committed 5 days ago

Verified 4b643c46

[Bugfix] Transformers backend: recompute `mm_token_type_ids` per request for M-RoPE (#46552)

decarpentierg committed 5 days ago

Verified 7544286b

[GLM5] Implement op fusion for GLM5/DSV3.2 (#46876)

WoosukKwon committed 5 days ago

Verified 89876b0c

[GLM5.2 Perf] Replace MOE all-reduce with reduce-scatter, 3.1%~3.2 E2E Throughput improvement (#46635)

yewentao256 committed 5 days ago

Verified 5c91039c

[ROCm][Perf][MLA] Add AITER FlashAttention MLA prefill backend (`ROCM_AITER_FA`) (#45033)

xaguilar-amd committed 5 days ago

Verified 5ecae326

[Bugfix][DSv3.2] Skip indexer weights for index-cache-skipped layers (#46600)

frida-andersson committed 6 days ago

Verified 6eb63a1d

[Render][Speculator] Add return_loss_mask to render endpoint for training data generation (#46846)

WindChimeRan committed 6 days ago

Verified 09841ae7

[Hardware][AMD][CI] Tweak mirrored tests; improve CI base dependency change detection (#46930)

mawong-amd committed 6 days ago

Verified a2a92cbb

[Bugfix][MM][CG] Enable dual-path ViT CUDA graph for Step3-VL (#46034)

shen-shanshan committed 6 days ago

Verified 35e6c86c

[ROCm][Perf] Add Fused Shared Expert (FSE) support for GLM-4.5/6/7 (#44313)

omirosh committed 6 days ago

Verified c7ca0bcc

Newer Older