As part of an effort to deepen my expertise in Rust programming and to refresh my knowledge of Large Language Models I have developed a Rust version of the Python version. The implementation utilizes the Burn library for tensor operations.

I have seen someone already posted the book's code using the Candle library but my project is a little bit less ambitious it only implements the training and generation in a tool that can be used to play with the model.

I believe that having multiple implementations of the model in different languages and frameworks opens up a fantastic opportunity for bench-marking and performance analysis. Comparing the speed, memory usage, and overall efficiency of these different approaches could provide valuable insights.

Thank you for your work