test: [DYN-1988] skip TRTLLM KVBM disagg test for now until upstream's fix is released (#5974)

tests/kvbm_integration/common.py

@@ -336,7 +336,7 @@ class DeterminismTester(ApiTester):
 
         top_k = -1
         if check_module_available("tensorrt_llm"):
-            top_k = 0
+            top_k = 1  # TensorRT-LLM requires top_k>=0 and dynamo frontend does not support top_k=0
         # For determinism: use temperature=0 which should trigger greedy decoding in vLLM
         # Setting top_p=1.0 and top_k=-1 to avoid any sampling/filtering
         return super().make_request(

tests/kvbm_integration/test_determinism_disagg.py

@@ -547,6 +547,10 @@ def tester(llm_server):
 class TestDeterminismDisagg(BaseTestDeterminism):
     """Test class for determinism validation."""
 
+    @pytest.mark.skipif(
+        check_module_available("tensorrt_llm"),
+        reason="Skipping test until the TRT-LLM disagg hang issue is fixed. (https://github.com/NVIDIA/TensorRT-LLM/pull/11247)",
+    )
     @pytest.mark.parametrize(
         "llm_server",
         [