Commits vllm-project/vllm

[WideEP] Remove naive all2all. Use allgather_reducescatter instead

tlrmchlsmth committed 22 days ago

6523da78

[MISC] Fix Tensor Parallelism for Quantized Mamba Models with n_groups=1 (#33257)

vadiklyutiy committed 22 days ago

Verified a372f3f4

Turn `@config` into a `dataclass_transform` (#31541)

hmellor committed 22 days ago

Verified 61e632ae

[torch.compile] Significantly speed up cold start times (#33641)

zou3519 committed 22 days ago

Verified b1bb18de

[Attention][FA3] Update FA3 to include new swizzle optimization (#23465)

LucasWilkinson committed 22 days ago

Verified 2267cb1c

[P/D] rework mooncake connector and introduce its bootstrap server (#31034)

dtcccc committed 22 days ago

Verified 0d6ccf68

[Bugfix] Fix startup hang for Granite Speech (#33699)

DarkLight1337 committed 22 days ago

Verified 18e7cbbb

[Voxtral models] Skip warm-up to skip confusing error message in warm-up (#33576)

patrickvonplaten committed 22 days ago

Verified f0d52517

[MM] Pass `prefix` parameter to MMEncoderAttention (#33674)

shen-shanshan committed 22 days ago

Verified 5c4f2dd6

[Bugfix] Do not add extra \n for image-only cases when constructing multimodal text prompts. (#33647)

noooop committed 22 days ago

Verified f3d8a346

Feat/add nemotron nano v3 tests (#33345)

shaharmor98 committed 22 days ago

Verified 4bc913ae

[Bugfix][Async][Connector] avoid vllm-side double free during async scheduling + request abort + async KV cache transfer (#33377)

KuntaiDu committed 22 days ago

Verified fbb3cf69

Document NixlConnector backend selection via kv_connector_extra_config (#33552)

KrxGu committed 22 days ago

Verified 2df2b349

Fix Gemma3n audio encoder for Transformers v5 (#33673)

hmellor committed 22 days ago

Verified 2a8d84e6

[Models] Intern-S1-Pro (#33636)

CUHKSZzxy committed 22 days ago

Verified a3acfa10

Fix Gemma3 GGUF for Transformers v5 (#33683)

hmellor committed 22 days ago

Verified be8168ff

Fix offline test for Transformers v5 (#33682)

hmellor committed 22 days ago

Verified f6af3462

[Bugfix] fix qwen3-asr response error (#33644)

jesse996 committed 22 days ago

Verified ceab70c8

[Misc] Update default image format of `encode_base64` (#33656)

DarkLight1337 committed 22 days ago

Verified 52683ccb

[Bugfix] Disable RoutingMethodType.[Renormalize,RenormalizeNaive] TRTLLM per-tensor FP8 MoE (#33620)

mgoin committed 23 days ago

Verified e346e2d0

[Refactor] Clean up pooling serial utils (#33665)

DarkLight1337 committed 23 days ago

Verified 83449a5f

[Bugfix][Model] Fix DeepSeek-OCR-2 chat template to include BOS token (#33642)

l4b4r4b4b4 committed 23 days ago

Verified dad2d6a5

[CI/Build] Investigate torchrun distributed tests hanging issue (#33650)

Isotr0py committed 23 days ago

Verified 32e84fa1

[torch.compile] Document the workaround to standalone_compile failing (#33571)

zou3519 committed 23 days ago

Verified fd9c83d0

[Misc] Remove deprecated VLLM_ALL2ALL_BACKEND environment variable (#33535)

carlory committed 23 days ago

Verified b95cc501

[Minor] Some code simplification in `scheduler.py` (#33597)

njhill committed 23 days ago

Verified 61397891

[Misc] Remove deprecated profiler environment variables (#33536)

carlory committed 23 days ago

Verified ef248ff7

[XPU][1/N] Deprecate ipex and switch to vllm-xpu-kernels for xpu platform (#33379)

jikunshang committed 23 days ago

Verified e1060448

[Bugfix] Interleaved thinking keeps compatibility with reasoning_content (#33635)

chaunceyjiang committed 23 days ago

Verified bf001da4

[CI/Build] Remove hardcoded America/Los_Angeles timezone from Dockerfiles (#33553)

carlory committed 23 days ago

Verified a0a984ac