vllm - 💡(How to fix) Fix SyntaxWarning: invalid escape sequence and spurious trust_remote_code warnings on startup

Bug Description

Two minor issues on every vLLM startup:

1. SyntaxWarning: invalid escape sequence \e

/opt/vllm-source/vllm/entrypoints/openai/chat_completion/protocol.py:346: SyntaxWarning: invalid escape sequence '\e'
  "(e.g. 'abcdabcdabcd...' or '\emoji \emoji \emoji ...'). This feature "
/opt/vllm-source/vllm/entrypoints/openai/completion/protocol.py:176: SyntaxWarning: invalid escape sequence '\e'
  "(e.g. 'abcdabcdabcd...' or '\emoji \emoji \emoji ...'). This feature "

The string literal contains \emoji which Python interprets as an invalid escape sequence. Should be either a raw string (r"...") or use \\emoji.

Files:

• vllm/entrypoints/openai/chat_completion/protocol.py:346
• vllm/entrypoints/openai/completion/protocol.py:176

2. Spurious trust_remote_code warning

(APIServer pid=1) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.
(APIServer pid=1) The argument `trust_remote_code` is to be used with Auto classes. It has no effect here and is ignored.

This warning is printed twice on every startup when --trust-remote-code is passed, even for models that DO use custom code (e.g., Qwen3-MoE). The warning comes from HuggingFace transformers when trust_remote_code is passed to a non-Auto class. vLLM should either suppress this warning or not pass the flag to classes that don't need it.

Environment

• vLLM version: 0.17.1 (via ghcr.io/llm-d/llm-d-cuda:v0.6.0)
• Python 3.12
• Model: RedHatAI/Qwen3-30B-A3B-FP8-dynamic (reproduces with any model)

Impact

Low severity — these are cosmetic warnings that don't affect functionality. However, they clutter logs and can confuse users running automated log analysis. The SyntaxWarning will become a SyntaxError in a future Python version.