vllm-project
diff --git a/‎vllm/v1/worker/gpu_model_runner.py‎
Lines changed: 1 addition & 1 deletion b/‎vllm/v1/worker/gpu_model_runner.py‎
Lines changed: 1 addition & 1 deletion
@@ -4000,7 +4000,7 @@ def _dummy_run(
  num_reqs=num_reqs_padded,
  max_query_len=max_query_len,
  ubatch_slices=ubatch_slices,
- for_cudagraph_capture=True,
+ for_cudagraph_capture=is_graph_capturing,
  )
 
  with self.maybe_dummy_run_with_lora(
Original file line number	Diff line number	Diff line change
`@@ -4000,7 +4000,7 @@ def _dummy_run(`
`4000`	`4000`	`num_reqs=num_reqs_padded,`
`4001`	`4001`	`max_query_len=max_query_len,`
`4002`	`4002`	`ubatch_slices=ubatch_slices,`
`4003`		`- for_cudagraph_capture=True,`
	`4003`	`+ for_cudagraph_capture=is_graph_capturing,`
`4004`	`4004`	`)`
`4005`	`4005`
`4006`	`4006`	`with self.maybe_dummy_run_with_lora(`