Nick Hill 44822d7ff2

[BugFix] Preserve spec decoding uniform decode when scheduling (#29759) ...

Signed-off-by: Nick Hill <nhill@redhat.com>

2025-12-01 17:15:52 -08:00

..

attention

[v1] Add real sliding window calculation to FlexAttention direct BlockMask building (#26015)

2025-12-01 13:12:51 +00:00

core

[BugFix] Preserve spec decoding uniform decode when scheduling (#29759)

2025-12-01 17:15:52 -08:00

engine

Make PyTorch profiler gzip and CUDA time dump configurable (#29568)

2025-12-01 04:30:46 +00:00

executor

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

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

kv_offload

[Attention] Update attention imports (#29540)

2025-11-27 11:19:09 -05:00

metrics

[Core][Observability] Add KV cache residency metrics (#27793)

2025-12-01 18:27:53 +00:00

pool

[mypy] Pass type checking for vllm/utils and vllm/v1/pool (#29666)

2025-11-28 20:40:47 +08:00

sample

[mypy] Enable type checking for more directories (#29674)

2025-11-28 08:39:27 -08:00

spec_decode

[Core] Enable inputs_embeds_size separate from hidden_size (#29741)

2025-11-30 17:31:12 +08:00

structured_output

[Misc] Refactor tokenizer interface (#29693)

2025-11-29 04:02:21 -08:00

worker

Make PyTorch profiler gzip and CUDA time dump configurable (#29568)

2025-12-01 04:30:46 +00:00

__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

[BugFix] Fix spec decoding max_tokens scheduling perf issue (#29542)

2025-11-28 20:52:23 +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