Yeah I really wish ollama didn't do that. I tried several ways to make it keep the model in memory like an environment variable but I've so far failed.
1. If using `ollama run`: `ollama run llama3 --keepalive -1`
2. If running ollama serve directly, use `OLLAMA_KEEP_ALIVE=-1` ollama serve
3. If using the api, there's a `keep_alive` parameter you can set to -1