[CI][ROCm] Restrict MLA cross-layer KV cache test to supported backends on ROCm (#46401)

Signed-off-by: aarushjain29 <Aarushi.Jain2@amd.com>

tests/v1/kv_connector/unit/test_kv_cache_layout.py

@@ -3,6 +3,8 @@
 
 import pytest
 
+from vllm.platforms import current_platform
+
 
 def test_mla_common_backend_rejects_cross_layer_kv_cache():
     """MLACommonBackend defaults to the identity permutation (layers dim
@@ -24,8 +26,13 @@ def test_mla_common_backend_rejects_cross_layer_kv_cache():
 
 @pytest.mark.parametrize(
     "backend_path",
+    # See: https://github.com/vllm-project/vllm/issues/46411
     [
         "vllm.v1.attention.backends.mla.triton_mla.TritonMLABackend",
+    ]
+    if current_platform.is_rocm()
+    else [
+        "vllm.v1.attention.backends.mla.triton_mla.TritonMLABackend",
         "vllm.v1.attention.backends.mla.cutlass_mla.CutlassMLABackend",
         "vllm.v1.attention.backends.mla.flashattn_mla.FlashAttnMLABackend",
         "vllm.v1.attention.backends.mla.flashmla.FlashMLABackend",