Starts with building micrograd to build an understanding of how pytorch understands how to calculate gradients, then it proceeds all the way to making a gpt-2 clone.
Looks like this is an effort to reorganize and build on that existing work.
