[Frontend] remove --device arg

Signed-off-by: Kebe <[email protected]>

Author: kebe7jun

docs/source/getting_started/installation/gpu/xpu.inc.md

@@ -66,7 +66,6 @@ XPU platform supports **tensor parallel** inference/serving and also supports **
 python -m vllm.entrypoints.openai.api_server \
      --model=facebook/opt-13b \
      --dtype=bfloat16 \
-     --device=xpu \
      --max_model_len=1024 \
      --distributed-executor-backend=ray \
      --pipeline-parallel-size=2 \

vllm/engine/arg_utils.py

@@ -19,7 +19,7 @@
 import vllm.envs as envs
 from vllm.config import (BlockSize, CacheConfig, CacheDType, CompilationConfig,
                          ConfigFormat, ConfigType, DecodingConfig,
-                         DetailedTraceModules, Device, DeviceConfig,
+                         DetailedTraceModules, DeviceConfig,
                          DistributedExecutorBackend, GuidedDecodingBackend,
                          GuidedDecodingBackendV1, HfOverrides, KVEventsConfig,
                          KVTransferConfig, LoadConfig, LoadFormat, LoRAConfig,
@@ -354,7 +354,6 @@ class EngineArgs:
     max_prompt_adapter_token: int = \
         PromptAdapterConfig.max_prompt_adapter_token
 
-    device: Device = DeviceConfig.device
     num_scheduler_steps: int = SchedulerConfig.num_scheduler_steps
     multi_step_stream_outputs: bool = SchedulerConfig.multi_step_stream_outputs
     ray_workers_use_nsight: bool = ParallelConfig.ray_workers_use_nsight
@@ -731,14 +730,6 @@ def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
             "--max-prompt-adapter-token",
             **prompt_adapter_kwargs["max_prompt_adapter_token"])
 
-        # Device arguments
-        device_kwargs = get_kwargs(DeviceConfig)
-        device_group = parser.add_argument_group(
-            title="DeviceConfig",
-            description=DeviceConfig.__doc__,
-        )
-        device_group.add_argument("--device", **device_kwargs["device"])
-
         # Speculative arguments
         speculative_group = parser.add_argument_group(
             title="SpeculativeConfig",
@@ -977,7 +968,7 @@ def create_engine_config(
         from vllm.platforms import current_platform
         current_platform.pre_register_and_update()
 
-        device_config = DeviceConfig(device=self.device)
+        device_config = DeviceConfig(device=current_platform.device_type)
         model_config = self.create_model_config()
 
         # * If VLLM_USE_V1 is unset, we enable V1 for "supported features"