You were able to do the same with StableLM before you commented here. It already has preliminary support in llama.cpp, which runs llama-7B in 5GB of RAM off of

You were able to do the same with StableLM before you commented here. It already has preliminary support in llama.cpp, which runs llama-7B in 5GB of RAM off of CPU faster than ChatGPT. https://github.com/ggerganov/llama.cpp/issues/1063#issuecomment-1515430933 https://github.com/ggerganov/llama.cpp/issues/1063#issuecomm...