Mohammad Miadh Angkad 61e381dcf0 [Perf] Add SM 10.3 (B300/GB300) all-reduce communicator tuning (#37756) ...

Signed-off-by: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com>

2026-03-21 19:43:47 +00:00

..

passes

[Perf] Add SM 10.3 (B300/GB300) all-reduce communicator tuning (#37756)

2026-03-21 19:43:47 +00:00

__init__.py

[torch.compile] avoid Dynamo guard evaluation overhead (#7898)

2024-08-28 16:10:12 -07:00

backends.py

Revert "[compile] Initialize passes at VllmBackend init" (#37733)

2026-03-20 21:35:49 -07:00

base_static_graph.py

Update Optional[x] -> x | None and Union[x, y] to x | y (#26633)

2025-10-12 09:51:31 -07:00

caching.py

[compile] Add compiled artifact counter for VLLM_USE_MEGA_AOT_ARTIFACT=1. (#37589)

2026-03-20 16:22:46 +00:00

compiler_interface.py

[Bugfix] Fix mock.patch resolution failure for standalone_compile.FakeTensorMode on Python <= 3.10 (#37158)

2026-03-17 20:13:06 +00:00

counter.py

[compile] aot_compile should respect VLLM_DISABLE_COMPILE_CACHE (#36358)

2026-03-11 03:12:03 -07:00

cuda_graph.py

[Perf] Fix slow hasattr in CUDAGraphWrapper.__getattr__ (#37425)

2026-03-19 15:43:48 +08:00

decorators.py

Revert "[compile] Initialize passes at VllmBackend init" (#37733)

2026-03-20 21:35:49 -07:00

monitor.py

[compile] Split compile/warmup monitoring (#36098)

2026-03-07 13:52:11 -08:00

partition_rules.py

[Misc][BE] Type coverage for vllm/compilation [1/3] (#31554)

2026-01-06 20:37:51 -05:00

piecewise_backend.py

[torch.compile] Use FakeTensors instead of real GPU tensors for single-size compilation (#36093)

2026-03-11 16:07:09 +00:00

wrapper.py

[torch.compile][BE][Multimodal] Remove requirement to set_model_tag to avoid cache conflict (#37345)

2026-03-19 17:26:12 +00:00