A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 7, 00:15 UTC
[Model Runner V2] Add dummy profile_cudagraph_memory API (#36520)
Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>
Woosuk Kwon committed
Mar 9, 2026 at 10:20 UTC
6e956d9eca398005929d29f123607d1029800cc7
1 file changed
+4
vllm/v1/worker/gpu/model_runner.py
+4
@@ -473,6 +473,10 @@ class GPUModelRunner(LoRAModelRunnerMixin):
473
# SP is not supported yet.
474
return num_scheduled_tokens
475
476
+ def profile_cudagraph_memory(self) -> int:
477
+ # NOTE(woosuk): It is TBD whether we keep this API or not.
478
+ return 0
479
+
480
@torch.inference_mode()
481
def capture_model(self) -> int:
482
if not self.cudagraph_manager.needs_capture():