vllm-project/vllm

Pull Requests Commits

Revert "[Bugfix] default set cuda_graph_sizes to max_num_seqs for v1 engine" (#20128)

mgoin committed 240 days ago

Verified 1f5d178e

[Bugfix][V1][ROCm] Fix AITER Flash Attention Backend (Fix API Break and Local Attention Logic: affecting Llama4) (#19904)

tjtanaa committed 240 days ago

Verified 27c065df

[Docs] Improve frameworks/helm.md (#20113)

windsonsea committed 240 days ago

Verified 84c260ca

[Misc] Use collapsible blocks for benchmark examples. (#20017)

reidliu41 committed 240 days ago

Verified 167aca45

[CPU] Fix torch version in x86 CPU backend (#19258)

bigPYJ1151 committed 240 days ago

Verified 0567c824

[Refactor] Remove unused library (#20099)

yewentao256 committed 240 days ago

Verified d188913d

[Doc] Update docs for New Model Implementation (#20115)

DarkLight1337 committed 240 days ago

Verified 1d7c29f5

[Bugfix] Allow `CUDA_VISIBLE_DEVICES=''` in `Platform.device_id_to_physical_device_id` (#18979)

eicherseiji committed 240 days ago

Verified 65397e40

[Benchmark][Bug] Fix multiple bugs in bench and add args to spec_decode offline (#20083)

ekagra-ranjan committed 240 days ago

Verified 9502c381

[PD] Skip `tp_size` exchange with rank0 (#19413)

NickLucche committed 240 days ago

Verified 25826835

[Bugfix] Fix Mistral tool-parser regex for nested JSON (#20093)

mgoin committed 240 days ago

Verified 754b00ed

[CI] Add SM120 to the Dockerfile (#19794)

mgoin committed 240 days ago

Verified 296ce95d

[TPU] Add TPU specific var VLLM_TPU_MOST_MODEL_LEN (#19919)

Chenyaaang committed 240 days ago

Verified 2d7620c3

[P/D] Avoid stranding blocks in P when aborted in D's waiting queue (#19223)

njhill committed 240 days ago

Verified 55c65ab4

[TPU][Bugfix] fix kv cache padding (#20048)

yaochengji committed 240 days ago

Verified 2cc20699

[Bugfix] default set cuda_graph_sizes to max_num_seqs for v1 engine (#20062)

izhuhaoran committed 240 days ago

Verified 9f0608fc

Fix the path to the testing script. (#20082)

QiliangCui committed 240 days ago

Verified 4e0db57f

[Misc] Add parallel state `node_count` function (#20045)

njhill committed 240 days ago

Verified c40692bf

[PD] let toy proxy handle /chat/completions (#19730)

lk-chen committed 241 days ago

Verified 4734704b

static_scaled_fp8_quant should not run when scale.numel is not 1 (#20076)

eldarkurtic committed 241 days ago

Verified 8b8c209e

[Fix] Support cls pooling in ModernBertPooler (#20067)

lsz05 committed 241 days ago

Verified 23a04e08

[Quantization] Add compressed-tensors emulations support for NVFP4 (#19879)

dsikka committed 241 days ago

Verified 02c97d9a

[Frontend] Add `/v1/audio/translations` OpenAI API endpoint (#19615)

NickLucche committed 241 days ago

Verified e795d723

[V1][Speculative Decoding] Fix DeepSeek MTP (#20022)

cjackal committed 241 days ago

Verified 8359f4c8

[Doc] Guide for Incremental Compilation Workflow (#19109)

mgoin committed 241 days ago

Verified bf518158

[doc] add reference link for Intel XPU (#20064)

reidliu41 committed 241 days ago

Verified c53fec1f

[BugFix] Fix full-cuda-graph illegal memory access in FA3 (#20057)

LucasWilkinson committed 241 days ago

Verified 0f9e7354

[Chore] debloat some initial logs (#19438)

aarnphm committed 241 days ago

Verified ba7ba35c

[Kernels][Bugfix] Use torch op for all kernels in FusedMoE forward. Add additional testing for cudagraphs. (#19717)

bnellnm committed 241 days ago

Verified 015fab8c

[Feat][CLI] enforce-include-usage (#19695)

max-wittig committed 241 days ago

Verified f59fc60f

Newer Older