From 6e956d9eca398005929d29f123607d1029800cc7 Mon Sep 17 00:00:00 2001 From: Woosuk Kwon Date: Mon, 9 Mar 2026 10:20:13 -0700 Subject: [PATCH] [Model Runner V2] Add dummy profile_cudagraph_memory API (#36520) Signed-off-by: Woosuk Kwon --- vllm/v1/worker/gpu/model_runner.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/vllm/v1/worker/gpu/model_runner.py b/vllm/v1/worker/gpu/model_runner.py index 8cb65c4d26..c26fe9d679 100644 --- a/vllm/v1/worker/gpu/model_runner.py +++ b/vllm/v1/worker/gpu/model_runner.py @@ -473,6 +473,10 @@ class GPUModelRunner(LoRAModelRunnerMixin): # SP is not supported yet. return num_scheduled_tokens + def profile_cudagraph_memory(self) -> int: + # NOTE(woosuk): It is TBD whether we keep this API or not. + return 0 + @torch.inference_mode() def capture_model(self) -> int: if not self.cudagraph_manager.needs_capture():