I trained a 115M model from scratch to use tools

In this video we show you how to train an AI model on the tiny stories dataset, a dataset that consists of childrens stories with no math capabilities. I show you the limitations of the model, and then we train the model in 2 different ways to learn math. the first version is training the model to do math via traditional techniques training the model on basic math. We then show the flaws of this approach. the second version is training the model to be able to call a calculator by creating a hole. this results in super accurate math, with even better story telling. repos -------------- vi