I think the original title was better: "Bridging the gap between neural networks and functions" It discusses the standard backpropagation optimization method i

I think the original title was better: "Bridging the gap between neural networks and functions" It discusses the standard backpropagation optimization method in differential form and the functional approximation of neural networks, but doesn't discuss transformers at all that I could tell. I think the code might be helpful to some in understanding implementation, but so much is now done in accelerators that it doesn't really capture real implementations.