Tags: #autograd

Deriving softmax, log-softmax, and NLL loss, then training an MLP on MNIST with a home-grown autograd.

How I scratched my itch for understanding gradients by building a simple autograd system