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

vllm/engine/arg_utils.py

@@ -504,12 +504,14 @@ class EngineArgs:
                   'Enabling this will use the fully sharded layers. '
                   'At high sequence length, max rank or '
                   'tensor parallel size, this is likely faster.'))
-        parser.add_argument(
-            "--device",
-            type=str,
-            default=EngineArgs.device,
-            choices=["auto", "cuda", "neuron", "cpu", "tpu", "xpu"],
-            help='Device type for vLLM execution.')
+        parser.add_argument("--device",
+                            type=str,
+                            default=EngineArgs.device,
+                            choices=[
+                                "auto", "cuda", "neuron", "cpu", "openvino",
+                                "tpu", "xpu"
+                            ],
+                            help='Device type for vLLM execution.')
 
         # Related to Vision-language models such as llava
         parser = EngineArgs.add_cli_args_for_vlm(parser)