vllm.compilation ¶

Modules:

• backends –
• base_static_graph –
• breakable_cudagraph –

  Breakable CUDA graph capture/replay.

• caching –
• codegen –

  Code generation for split_gm stitching graph execution.

• compiler_interface –
• cuda_graph –
• decorators –
• monitor –
• partition_rules –
• passes –
• piecewise_backend –
• wrapper –