Commits vllm-project/vllm

Merge branch 'main' into k3-vit-optimization

jeejeelee committed 67 days ago

Verified 4dbeac4a

[Bugfix][MLA] Fix BLHNC addressing for FlashInfer sparse MLA (#54465)

LucasWilkinson committed 67 days ago

Verified 57073552

[Bugfix] Make metadata send non-blocking in GroupCoordinator.isend_tensor_dict (#49274)

RookieCoder-Camera committed 67 days ago

Verified c92b29a1

[Flashinfer] Upgrade Flashinfer version to 0.6.18 (#54313)

wzhao18 committed 67 days ago

Verified 7ab29234

[CI] Restore gpu_1_queue routing for torch-abi audit (#54468)

khluu committed 67 days ago

Verified 7a100bb6

[Bugfix][Spec Decode] Keep default CUDA graph sizes memory-safe (#54418)

khluu committed 67 days ago

Verified b2dc864b

[Bugfix][Quantization][MoE] Route weight only NVFP4 checkpoints through W4A16 (#54427)

ima-helikoptaaa committed 67 days ago

Verified 9d0fe9ba

[codeowners] Add jperezdealgaba to security file ownership (#54358)

jperezdealgaba committed 67 days ago

Verified e79961c9

[Bugfix][Multimodal] Avoid caching full prompts in fallback (#54439)

waizuichougou committed 67 days ago

Verified f6895a5f

[Bugfix][Kernel] Keep packed GDN decode beta in FP32 (#53877)

CherryLemon committed 67 days ago

Verified 56058fd5

[Test][VLM] Add batch-invariance tests for Qwen3-VL (#53531)

machero committed 67 days ago

Verified 87b9b5b8

[Bugfix] Validate scale-out transfer params (#54324)

taneem-ibrahim committed 67 days ago

Verified fe6db3ed

[Doc] Fix griffe warnings in HYV4 tool parser (#54412)

hmellor committed 67 days ago

Verified 488b6da1

[CI] Mark L4 GPU test steps with device: l4 for EKS migration (#54326)

khluu committed 68 days ago

Verified 5e71a11e

ci: add MIG slice size to H200 job labels (#54420)

khluu committed 68 days ago

Verified 78fa1891

[Bugfix] Avoid global config lookup in sparse indexer forward (#54400)

ZJY0516 committed 68 days ago

Verified 8c51b926

[Perf][ROCm] Dual-stream decode with hipgraphs (#52033)

simondanielsson committed 68 days ago

Verified 2c7d7dd6

[Model] Honor cap_pixels_per_frame in Qwen3-VL memory profiling (#54380)

dkrisman committed 68 days ago

Verified 4f78a8fd

[Bugfix] Reset cached Mamba align metadata on profiling teardown (#54044)

gcanlin committed 68 days ago

Verified b383e163

[CI] Add explicit step keys to 18 hardware test steps (#54330)

khluu committed 68 days ago

Verified 8fa4c6cd

[Bugfix] Bound cache_salt length to prevent DoS via scheduler CPU exhaustion (#54353)

jperezdealgaba committed 68 days ago

Verified 1dc464d4

[Fix] Improve ROCm detection in WSL environments (#38434)

yiz-liu committed 68 days ago

Verified 1ebff996

[Rust Frontend][gRPC] Add audio and video media inputs (#53760)

Connor Carpenter committed 68 days ago

Verified 7a67941c

[Bugfix][Multimodal] Release Qwen2.5-VL and Qwen3-VL RoPE caches with the model (#54346)

waizuichougou committed 68 days ago

Verified b016ed8e

[ROCm][MLA] Reach FULL cudagraphs for AITER MLA speculative decoding (#51171)

yudigege86 committed 68 days ago

Verified dbf662c9

[Bugfix][MoE] Enable cuBLAS out_dtype router GEMM on all CUDA archs (fixes family-120/GB10) (#54048)

hclsys committed 68 days ago

Verified b5707bf9

[CI][Ray] Fix flaky multi-node assignment test after placement-group teardown (#53621)

yzong-rh committed 68 days ago

Verified 680e2177

[CI][Test] Deflake test_mem.py sleep-mode asserts via allocator bookeeping (#54312)

okorzh-amd committed 68 days ago

Verified 6cddad41

[Bugfix][Model Runner V2][Spec Decode] Decouple the draft's gumbel noise stream from the target's (#54282)

TheEpicDolphin committed 68 days ago

Verified fe755c88

[Bugfix][Frontend] Only echo the assistant turn in batched chat completions (#52529)

Kaif10 committed 68 days ago

Verified 7fbfca26