[https://nvbugs/5674665][chore] Add test coverage for https://nvbugspro.nvidia.com/bug/5674665 (#9518)

Signed-off-by: eopXD <yuehtingc@nvidia.com>

Author: eopXD

tests/integration/defs/accuracy/test_llm_api_pytorch.py

@@ -1104,6 +1104,41 @@ def test_auto_dtype_vswa_reuse(self):
             task = MMLU(self.MODEL_NAME)
             task.evaluate(llm)
 
+    def test_auto_dtype_vswa_without_reuse_disable_overlap_scheduler(self):
+        # NOTE: Test with VSWA kv cache config.
+        kv_cache_config = KvCacheConfig(
+            enable_block_reuse=False,
+            enable_partial_reuse=False,
+            max_attention_window=[512, 512, 512, 512, 512, 32768],
+        )
+
+        with LLM(self.MODEL_PATH,
+                 kv_cache_config=kv_cache_config,
+                 disable_overlap_scheduler=True) as llm:
+            task = GSM8K(self.MODEL_NAME)
+            task.evaluate(llm)
+            task = MMLU(self.MODEL_NAME)
+            task.evaluate(llm)
+
+    @pytest.mark.skip(
+        reason=
+        "Currently failing due to accuracy drop, https://nvbugspro.nvidia.com/bug/5674665"
+    )
+    def test_auto_dtype_vswa_reuse_disable_overlap_scheduler(self):
+        # NOTE: Test with VSWA kv cache config.
+        kv_cache_config = KvCacheConfig(
+            enable_block_reuse=True,
+            max_attention_window=[512, 512, 512, 512, 512, 32768],
+        )
+
+        with LLM(self.MODEL_PATH,
+                 kv_cache_config=kv_cache_config,
+                 disable_overlap_scheduler=True) as llm:
+            task = GSM8K(self.MODEL_NAME)
+            task.evaluate(llm)
+            task = MMLU(self.MODEL_NAME)
+            task.evaluate(llm)
+
     def test_auto_dtype_vswa_reuse_partial_reuse(self):
         # NOTE: Test with VSWA kv cache config.
         kv_cache_config = KvCacheConfig(

tests/integration/test_lists/test-db/l0_h100.yml

@@ -45,6 +45,7 @@ l0_h100:
   - accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_reuse_partial_reuse
   - accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_reuse_low_memory_available_no_partial_reuse
   - accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_reuse_low_memory_available_partial_reuse
+  - accuracy/test_llm_api_pytorch.py::TestGemma3_1BInstruct::test_auto_dtype_vswa_without_reuse_disable_overlap_scheduler
   - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16[attn_backend=TRTLLM-torch_compile=False]
   - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_bfloat16[attn_backend=TRTLLM-torch_compile=True]
   - accuracy/test_llm_api_pytorch.py::TestLlama3_1_8BInstruct::test_chunked_prefill[attn_backend=TRTLLM] TIMEOUT (90)