diff --git a/vllm/model_executor/layers/fla/ops/utils.py b/vllm/model_executor/layers/fla/ops/utils.py
index 5a48e56a5..18e17a511 100644
--- a/vllm/model_executor/layers/fla/ops/utils.py
+++ b/vllm/model_executor/layers/fla/ops/utils.py
@@ -119,7 +119,7 @@ def input_guard(fn: Callable[..., torch.Tensor]) -> Callable[..., torch.Tensor]:
 def get_available_device() -> str:
     try:
         return triton.runtime.driver.active.get_current_target().backend
-    except BaseException:
+    except (RuntimeError, AttributeError):
         return "cpu"