[Misc] Move device options to a single place (#8322)

benchmarks/benchmark_latency.py

@@ -10,7 +10,7 @@ import torch
 from tqdm import tqdm
 
 from vllm import LLM, SamplingParams
-from vllm.engine.arg_utils import EngineArgs
+from vllm.engine.arg_utils import DEVICE_OPTIONS, EngineArgs
 from vllm.inputs import PromptInputs
 from vllm.model_executor.layers.quantization import QUANTIZATION_METHODS
 from vllm.utils import FlexibleArgumentParser
@@ -205,13 +205,11 @@ if __name__ == '__main__':
         default=None,
         help=('path to save the pytorch profiler output. Can be visualized '
               'with ui.perfetto.dev or Tensorboard.'))
-    parser.add_argument(
-        "--device",
-        type=str,
-        default="auto",
-        choices=["auto", "cuda", "cpu", "openvino", "tpu", "xpu"],
-        help='device type for vLLM execution, supporting CUDA, OpenVINO and '
-        'CPU.')
+    parser.add_argument("--device",
+                        type=str,
+                        default="auto",
+                        choices=DEVICE_OPTIONS,
+                        help='device type for vLLM execution')
     parser.add_argument('--block-size',
                         type=int,
                         default=16,

benchmarks/benchmark_throughput.py

@@ -11,7 +11,7 @@ from tqdm import tqdm
 from transformers import (AutoModelForCausalLM, AutoTokenizer,
                           PreTrainedTokenizerBase)
 
-from vllm.engine.arg_utils import AsyncEngineArgs, EngineArgs
+from vllm.engine.arg_utils import DEVICE_OPTIONS, AsyncEngineArgs, EngineArgs
 from vllm.entrypoints.openai.api_server import (
     build_async_engine_client_from_engine_args)
 from vllm.model_executor.layers.quantization import QUANTIZATION_METHODS
@@ -451,13 +451,11 @@ if __name__ == "__main__":
         'accuracy issues. FP8_E5M2 (without scaling) is only supported on '
         'cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is '
         'instead supported for common inference criteria.')
-    parser.add_argument(
-        "--device",
-        type=str,
-        default="auto",
-        choices=["auto", "cuda", "cpu", "openvino", "tpu", "xpu"],
-        help='device type for vLLM execution, supporting CUDA, OpenVINO and '
-        'CPU.')
+    parser.add_argument("--device",
+                        type=str,
+                        default="auto",
+                        choices=DEVICE_OPTIONS,
+                        help='device type for vLLM execution')
     parser.add_argument(
         "--num-scheduler-steps",
         type=int,

vllm/engine/arg_utils.py

@@ -26,6 +26,16 @@ logger = init_logger(__name__)
 
 ALLOWED_DETAILED_TRACE_MODULES = ["model", "worker", "all"]
 
+DEVICE_OPTIONS = [
+    "auto",
+    "cuda",
+    "neuron",
+    "cpu",
+    "openvino",
+    "tpu",
+    "xpu",
+]
+
 
 def nullable_str(val: str):
     if not val or val == "None":
@@ -553,10 +563,7 @@ class EngineArgs:
         parser.add_argument("--device",
                             type=str,
                             default=EngineArgs.device,
-                            choices=[
-                                "auto", "cuda", "neuron", "cpu", "openvino",
-                                "tpu", "xpu"
-                            ],
+                            choices=DEVICE_OPTIONS,
                             help='Device type for vLLM execution.')
         parser.add_argument('--num-scheduler-steps',
                             type=int,