[Hardware][Intel] OpenVINO vLLM backend (#5379)

benchmarks/benchmark_throughput.py

@@ -349,9 +349,10 @@ if __name__ == "__main__":
     parser.add_argument(
         "--device",
         type=str,
-        default="cuda",
-        choices=["cuda", "cpu", "tpu", "xpu"],
-        help='device type for vLLM execution, supporting CUDA and CPU.')
+        default="auto",
+        choices=["auto", "cuda", "cpu", "openvino", "tpu", "xpu"],
+        help='device type for vLLM execution, supporting CUDA, OpenVINO and '
+        'CPU.')
     parser.add_argument(
         "--enable-prefix-caching",
         action='store_true',