[XPU] Fix decode graph runner is_current_stream_capturing on non-CUDA devices (#35050)

2026-08-18 14:28:13 +08:00
parent 27596abdc0
commit 5eb117f6ca
2 changed files with 4 additions and 1 deletions
@@ -467,9 +467,11 @@ class DecodeCudaGraphRunner(BaseCudaGraphRunner):
    def _record_in_graph_metadata_prep_done(self):
        # Purely a marker at this point in the graph; where the shared reads
        # actually end is the attn backend's call.
-        if not torch.cuda.is_current_stream_capturing():
+        if not self.device_module.is_current_stream_capturing():
            # Warmup shares this body. Breakable capture still plants: it opens
            # segment 1 on context entry and every segment re-arms the node.
+            # Routed through device_module so XPU (torch.xpu) is picked up
+            # instead of hitting torch.cuda dummy stubs on non-CUDA builds.
            return
        if self.in_graph_metadata_prep_done is None:
            self.in_graph_metadata_prep_done = make_external_event(self.device_module)