Some advanced confusion about LocalAI

[zenyanbo]

Hello, I really like LocalAI. However, I encountered some confusion while using it. LocalAI is composed of different backends, and different backends have different parameters and even usage methods. When I want to customize the settings further (for example, llama.cpp has richer parameters), how should I set it up?Also, exllama's debug messages don't seem to appear.

My idea is to understand how the parameters in the config are converted into backend commands, and then they can be set according to the documentation of llama.cpp, vllm, and exllama.

Therefore, I think I need to further understand the structure of LocalAI, but I am a newbie. Can anyone give me a suitable starting point?