Isotr0py 02080179a3 [Bugfix] Fix torchrun PP broadcast deadlock with async scheduling (#33701) ...

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>

2026-02-04 02:17:37 +00:00

..

attention

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

2026-02-03 08:08:47 -08:00

core

[BugFix][Spec Decoding] Fix negative accepted tokens metric crash (#33729)

2026-02-03 23:34:41 +00:00

engine

[Refactor] Move profiling methods to MM budget (#33559)

2026-02-02 23:27:00 +08:00

executor

Support clear mm and encoder cache (#33452)

2026-01-31 15:22:25 +00:00

kv_offload

[Metrics] [KVConnector] Add Offloading Connector metrics (#27942)

2026-01-27 13:34:49 +00:00

metrics

Support clear mm and encoder cache (#33452)

2026-01-31 15:22:25 +00:00

pool

[Refactor] Clean up pooler modules (#31897)

2026-01-08 00:07:43 +08:00

sample

[perf] v1/spec_decode: skip softmax for all-greedy rejection sampling (#32852)

2026-01-31 09:51:26 +00:00

spec_decode

Turn @config into a dataclass_transform (#31541)

2026-02-03 17:40:59 +00:00

structured_output

[code clean] remove duplicate code (#33135)

2026-01-27 04:57:16 +00:00

worker

[Bugfix] Fix torchrun PP broadcast deadlock with async scheduling (#33701)

2026-02-04 02:17:37 +00:00

__init__.py

[V1] AsyncLLM Implementation (#9826)

2024-11-11 23:05:38 +00:00

cudagraph_dispatcher.py

Reduce the kernel overhead when num of active loras is smaller than max loras. Multiple cuda graphs are captured for each num of active-loras. (#32005)

2026-02-02 12:30:06 -05:00

kv_cache_interface.py

Make mypy opt-out instead of opt-in (#33205)

2026-01-29 09:12:26 +00:00

outputs.py

[ModelRunner V2] Fix spec decoding + logprobs (#33391)

2026-01-31 03:33:26 +00:00

request.py

[Misc] Replace Optional[X] with X | None syntax (#33332)

2026-01-30 01:56:59 -08:00

serial_utils.py

[Multimodal] Simplify MM input definitions (#33331)

2026-01-29 13:32:04 +00:00

utils.py

[Misc] Replace Optional[X] with X | None syntax (#33332)

2026-01-30 01:56:59 -08:00