Wei Zhao b37b679770 [Feature][Perf] Support Selective CPU Weight Offloading (#34535) ...

Signed-off-by: wzhao18 <wzhao18.sz@gmail.com>

2026-02-13 20:02:24 -08:00

..

attention

[GDN] Use CPU tensors to build GDN metadata (#34498)

2026-02-13 01:24:45 -08:00

core

[Core] Move pause and resume functions into engine (#34125)

2026-02-13 00:15:10 -08:00

engine

[Core] Move pause and resume functions into engine (#34125)

2026-02-13 00:15:10 -08:00

executor

[Core] Profiler improvements and lazy initialization (#33198)

2026-02-12 16:16:38 -08:00

kv_offload

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

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

metrics

[Core] Profiler improvements and lazy initialization (#33198)

2026-02-12 16:16:38 -08:00

pool

[Refactor] Clean up pooler modules (#31897)

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

sample

Fix num_logprobs parameter description in sampler.py (#34451)

2026-02-12 18:18:03 -08:00

spec_decode

[Bugfix] Fix MTP accuracy for GLM-5 (#34385)

2026-02-12 11:08:19 +08:00

structured_output

[Refactor] Simplify BOS/EOS token handling (#34435)

2026-02-12 18:18:24 -08:00

worker

[Feature][Perf] Support Selective CPU Weight Offloading (#34535)

2026-02-13 20:02:24 -08:00

__init__.py

[V1] AsyncLLM Implementation (#9826)

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

cudagraph_dispatcher.py

Reapply [Attention][FA3] Update FA3 to include new swizzle optimization (#34043)

2026-02-11 07:07:56 -08:00

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

[Refactor] Simplify BOS/EOS token handling (#34435)

2026-02-12 18:18:24 -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