fix w/ cuda graph

Signed-off-by: Enwei Zhu <21126786+syuoni@users.noreply.github.com>

Author: syuoni

tensorrt_llm/_torch/pyexecutor/cuda_graph_runner.py

@@ -33,6 +33,7 @@ def __init__(
         device: str,
         attn_metadata: AttentionMetadata,
         spec_metadata: Optional[SpecMetadata] = None,
+        guided_metadata=None,
         use_mrope: bool = False,
         max_beam_width: int = 1,
     ) -> None:
@@ -71,6 +72,7 @@ def __init__(
 
         self.attn_metadata = attn_metadata
         self.spec_metadata = spec_metadata
+        self.guided_metadata = guided_metadata
         self._output = None
         self._graph = None
         self.optional_extra_model_inputs = ["mrope_position_deltas"]
@@ -90,6 +92,7 @@ def capture(
             "position_ids": self.position_ids,
             "inputs_embeds": None,
             "spec_metadata": self.spec_metadata,
+            "guided_metadata": self.guided_metadata,
             "mrope_position_deltas": self.mrope_position_deltas,
         }