GitLab

feed responses back into the llm

ggml-metal.metal must be in the same directory as the ollama binary
otherwise llama.cpp will not be able to find it and load it.

1. go generate llama/llama_metal.go
2. go build .
3. ./ollama serve