1 article tagged configuration.
Every one of the twelve models on this rig serves through the same OpenAI-compatible endpoint, and every one wanted something different before it would run. Here are my working settings for each - and, more useful than the flags, what the deal was with each model and how I got it serving.