diff --git a/tests/integration/defs/accuracy/test_llm_api_pytorch.py b/tests/integration/defs/accuracy/test_llm_api_pytorch.py index ad192d34d9a5..b7b774e48911 100644 --- a/tests/integration/defs/accuracy/test_llm_api_pytorch.py +++ b/tests/integration/defs/accuracy/test_llm_api_pytorch.py @@ -7115,6 +7115,7 @@ def test_fp8_4gpus(self, attention_dp, use_cpp_mamba, monkeypatch): enable_block_reuse=False, mamba_ssm_cache_dtype="float16", free_gpu_memory_fraction=0.5, + use_kv_cache_manager_v2=False, ), max_batch_size=32, tensor_parallel_size=4,