The community member wants to build a RAG using LlamaIndex and Llama3.1 without having to install and download the model every time from Ollama. Another community member suggests that Ollama downloads the model weights only the first time and stores them in the ~/.ollama/models directory. The community member can modify the path by updating the OLLAMA_MODELS environment variable to a custom location, and the model will be downloaded there the first time and reused in the future.
I want to build a RAG using LlamaIndex and Llama3.1 but I don't want to install and download the model every time from ollama. Is there a way to download Llama3.1 and then load it to be used with the LlamaIndex framework?
Ollama downloads model weights only the first time and loads them later on. the models are stored in ~/.ollama/models. You can modify the path by updating OLLAMA_MODELS env variable: OLLAMA_MODELS='<path>'. Then the first time, it'll be downloaded there and the weights will be reused in the future.
