diff --git a/tests/models/decoder_only/vision_language/test_models.py b/tests/models/decoder_only/vision_language/test_models.py
index 163752e9fe06e..1ab42f8c126f8 100644
--- a/tests/models/decoder_only/vision_language/test_models.py
+++ b/tests/models/decoder_only/vision_language/test_models.py
@@ -34,6 +34,7 @@
     "dtype": "half",
     "max_tokens": 5,
     "tensor_parallel_size": 2,
+    "model_kwargs": {"device_map": "auto"},
     "image_size_factors": [(.25, 0.5, 1.0)],
     "distributed_executor_backend": (
         "ray",
diff --git a/tests/models/decoder_only/vision_language/vlm_utils/types.py b/tests/models/decoder_only/vision_language/vlm_utils/types.py
index fd18c7c8346f0..8459476dc2d07 100644
--- a/tests/models/decoder_only/vision_language/vlm_utils/types.py
+++ b/tests/models/decoder_only/vision_language/vlm_utils/types.py
@@ -158,6 +158,7 @@ def get_non_parametrized_runner_kwargs(self):
             "max_model_len": self.max_model_len,
             "max_num_seqs": self.max_num_seqs,
             "task": self.task,
+            "tensor_parallel_size": self.tensor_parallel_size,
             "hf_output_post_proc": self.hf_output_post_proc,
             "vllm_output_post_proc": self.vllm_output_post_proc,
             "auto_cls": self.auto_cls,