haosdent 0191354827 [Perf][MLA] Enable FULL cudagraph capture for TRITON_MLA decode (#42885) ...

Signed-off-by: haosdent <haosdent@gmail.com>

2026-05-18 14:29:10 -07:00

..

attention

[Perf][MLA] Enable FULL cudagraph capture for TRITON_MLA decode (#42885)

2026-05-18 14:29:10 -07:00

core

[Core][DSV4] Skip caching SWA blocks that can never serve a prefix-cache hit (#42258)

2026-05-15 15:59:18 +08:00

engine

[V1][DP][LB] Publish request counts at the start of each engine step (#41626)

2026-05-14 15:39:24 +00:00

executor

fix nixl side-channel host selection (#41806)

2026-05-11 07:40:37 +00:00

kv_offload

Tier offload followup (#42529)

2026-05-18 19:41:58 +00:00

metrics

[Bugfix][Metrics] Fix RayPrometheusMetric.labels() returning shared labeled child (#40840)

2026-05-01 08:43:39 +01:00

pool

[chore] Refactor pooling metadata token ID accessors (#42368)

2026-05-13 06:08:01 +00:00

sample

[XPU] keep generator state of sycl kernel align with pytorch (#41771)

2026-05-12 11:44:47 +00:00

simple_kv_offload

[Bugfix][SimpleCPUOffloadBackend] Dedup in-flight CPU offload stores across scheduler steps (#41289)

2026-05-13 01:53:32 -07:00

spec_decode

[Bugfix] DFlash FP8 KV-Cache (#42692)

2026-05-15 08:29:45 -06:00

structured_output

[Bugfix] Pass reasoning parser kwargs to structured output (#41199)

2026-04-30 23:38:02 -07:00

worker

[Model Runner V2] Fix prompt logprobs calculation Sizes of tensors must match error (#42778)

2026-05-18 08:27:21 -07:00

__init__.py

[V1] AsyncLLM Implementation (#9826)

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

cudagraph_dispatcher.py

[Experimental] Breakable CUDA graph (#42304)

2026-05-16 22:04:12 +08:00

kv_cache_interface.py

[Bugfix][Hybrid][NemotronH] Fix mamba_cache_mode=all + speculative decoding crash (#41233)

2026-05-18 14:54:00 +03:00

outputs.py

[RFC] Replace shared-memory routed experts with ModelRunnerOutput transfer and HTTP support (#39568)

2026-05-14 14:12:30 +00:00

request.py

[Bugfix][KV Transfer][NIXL] Notify P node on pre-admission rejection to free stranded KV blocks (#41269)

2026-05-09 22:52:09 -07:00

serial_utils.py

Add tensor IPC transfer mechanism for multimodal data (#32104)

2026-03-21 20:10:20 +00:00

utils.py

[Refactor] Unify engine process monitoring in engine manager and add Ray backend support (#35862)

2026-03-30 10:16:09 -07:00