Woosuk Kwon 11ea5ec1ff

[Model Runner V2] Refactor CudaGraphManager (#29583) ...

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

2025-11-26 21:37:59 -08:00

..

attention

[Attention][Async] Eliminate seq_lens_cpu in FlashAttention metadata building with DCP > 1 (#29449)

2025-11-26 18:48:43 -08:00

core

[Core] Support logprobs with spec decode + async scheduling (#29223)

2025-11-25 12:55:24 -08:00

engine

Scheduled removal of override_pooler_config and disable_log_requests (#29402)

2025-11-25 16:08:57 +00:00

executor

[Feature] Prefill Context Parallel (PCP) basic support (#28718)

2025-11-19 15:52:44 -05:00

kv_offload

[Attention] Remove imports from vllm/attention/__init__.py (#29342)

2025-11-26 10:53:15 -07:00

metrics

[Metrics] Scheduled removal of deprecated metrics (#29330)

2025-11-25 11:15:13 +08:00

pool

…

sample

[Core] Support logprobs with spec decode + async scheduling (#29223)

2025-11-25 12:55:24 -08:00

spec_decode

Revert back to torch.equal over torch.allclose from #28819 (#29086)

2025-11-25 14:23:38 +00:00

structured_output

[Bugfix] Fix llguidance backend, rollback when EOS was encountered (#25905)

2025-11-10 14:53:32 -05:00

worker

[Model Runner V2] Refactor CudaGraphManager (#29583)

2025-11-26 21:37:59 -08:00

__init__.py

…

cudagraph_dispatcher.py

[Core] Refactor padding logic and pad for CUDA graphs before attention metadata building (#28579)

2025-11-26 14:07:13 -05:00

kv_cache_interface.py

[Feature] Prefill Context Parallel (PCP) basic support (#28718)

2025-11-19 15:52:44 -05:00

outputs.py

Revert "[Redo] #26368 (#28771)" (#29121)

2025-11-20 21:27:45 -08:00

request.py

[KV Connector] Fix async connector prefix cache metrics (#28585)

2025-11-21 17:45:00 -05:00

serial_utils.py

[DisaggEverything] Tokens in<>out /generate endpoint (#24261)

2025-11-14 09:58:01 -07:00

utils.py

[BugFix] Fix async scheduling + chunked prefill + preemption (#28787)

2025-11-17 06:46:46 +08:00