[NPU] Support radix-cache with mamba-extra-buffer for Qwen3.5 (#23891)

2026-05-11 09:40:44 +08:00
parent e9a15b95da
commit 0fc9598b3e
1 changed files with 2 additions and 2 deletions
@@ -2465,8 +2465,8 @@ class ServerArgs:
                )

            assert (
-                is_cuda() or is_musa()
-            ), "Mamba extra_buffer is only supported on CUDA and MUSA devices with FLA backend"
+                is_cuda() or is_musa() or is_npu()
+            ), "Mamba extra_buffer is only supported on CUDA and MUSA and NPU devices with FLA backend"
            if self.speculative_num_draft_tokens is not None:
                assert (
                    self.mamba_track_interval >= self.speculative_num_draft_tokens