https://www.reddit.com/r/LocalLLaMA/comments/18hq98b/comment/kdaytgv/?utm_source=share&utm_medium=web2x&context=3
The reasoning is, if you already have a functional LLM setup, you shouldn't need to compile from scratch (which can be super annoying if you are eg on CUDA)
https://www.reddit.com/r/LocalLLaMA/comments/18hq98b/comment/kdaytgv/?utm_source=share&utm_medium=web2x&context=3
The reasoning is, if you already have a functional LLM setup, you shouldn't need to compile from scratch (which can be super annoying if you are eg on CUDA)