Fix optional graph_signature access in memory planning (pytorch#9325)

### Summary

CI is broken due to pytorch#8031.
`graph_signature` is optional, and is None.

### Test plan

CI

cc @JacobSzwejbka @angelayi

Author: jathu

exir/memory_planning.py

@@ -248,16 +248,7 @@ def verify_graph_input_output(self) -> None:
                     has_dynamic_unbound_output |= has_dynamic_unbound_tensor
 
         # only check if inputs are allocated if there are user inputs:
-        user_inputs_exist = (
-            len(
-                list(
-                    filter(
-                        lambda input: input.kind == InputKind.USER_INPUT,
-                        self.graph_signature.input_specs,
-                    )
-                )
-            )
-        ) > 0
+        user_inputs_exist = _do_user_inputs_exist(graph_signature=self.graph_signature)
 
         if "placeholder" in check_list and user_inputs_exist:
             assert graph_input_allocated is not None, "graph_input_allocated not set"
@@ -339,6 +330,22 @@ def _is_mutable_buffer(
     return False
 
 
+def _do_user_inputs_exist(graph_signature: Optional[ExportGraphSignature]) -> bool:
+    if graph_signature is None:
+        return False
+
+    return (
+        len(
+            list(
+                filter(
+                    lambda input: input.kind == InputKind.USER_INPUT,
+                    graph_signature.input_specs,
+                )
+            )
+        )
+    ) > 0
+
+
 def get_graph_input_tensors(
     nodes: Iterable[Node], graph_signature: Optional[ExportGraphSignature] = None
 ) -> Set[TensorSpec]: