Someone should try this on that model: https://www.oxen.ai/blog/training-a-rust-1-5b-coder-lm-with-...