Configuration needed
#24
by
vladciocan88
- opened
Hello , thank you again for a great model <3
I'm currently running this model in vLLM following the official documentation, but the quality of the responses seems noticeably weaker compared to the demo version. Could you share the recommended or optimal inference configuration/parameters (e.g., decoding settings, temperature, top-p, etc.) to achieve similar output quality to the demo?