vllm - 💡(How to fix) Fix [Feature]: Support logging the complete request prompt and model output to the log for easier debugging. [1 comments, 2 participants]

TL;DR

Implement logging for the complete request prompt and model output to facilitate easier debugging.

Guidance

• Review the documentation at https://docs.vllm.ai/en/latest/ for potential logging configuration options or examples.
• Search existing issues for similar feature requests or logging implementations that could serve as a reference.
• Consider reaching out to the community or the chatbot mentioned for guidance on how to implement the desired logging functionality.
• Evaluate the current logging setup to identify where modifications can be made to include the request prompt and model output.

Notes

The provided information lacks specific technical details about the current implementation, making it challenging to offer a precise solution.

Recommendation

Apply workaround: Implement a custom logging solution to capture the required information, as the existing documentation and resources may not provide a direct, built-in method for logging the complete request prompt and model output.