My #RAG on sbc isn't going great, I feel let down, it's not about the programming is mostly about the hardware limitations of the SBC - 4 GB #AIPU doesn't seem enough - 16 GB AIPU should be the right amount!

Also there is the llm limitation, I cannot use IBM's #granite because the voyager sdk is not supporting it! So I only have only choise for now, to use llama 3b from over 1year ago! Also I cannot use the embedding model on AIPU and on sbc's cpu works so slow.. I can mitigate this tho...