[Bugfix] set VLLM_WORKER_MULTIPROC_METHOD=spawn for vllm.entrypoionts.openai.api_server (#15700)

Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com>

vllm/entrypoints/openai/api_server.py

@@ -82,7 +82,8 @@ from vllm.entrypoints.openai.serving_tokenization import (
 from vllm.entrypoints.openai.serving_transcription import (
     OpenAIServingTranscription)
 from vllm.entrypoints.openai.tool_parsers import ToolParserManager
-from vllm.entrypoints.utils import load_aware_call, with_cancellation
+from vllm.entrypoints.utils import (cli_env_setup, load_aware_call,
+                                    with_cancellation)
 from vllm.logger import init_logger
 from vllm.reasoning import ReasoningParserManager
 from vllm.transformers_utils.config import (
@@ -1106,6 +1107,7 @@ if __name__ == "__main__":
     # NOTE(simon):
     # This section should be in sync with vllm/entrypoints/cli/main.py for CLI
     # entrypoints.
+    cli_env_setup()
     parser = FlexibleArgumentParser(
         description="vLLM OpenAI-Compatible RESTful API server.")
     parser = make_arg_parser(parser)