(WIP) My own implementation of Karpathy's micrograd. There are no differences so far, I hope there will be some soon! UPDATE: There actually is now! I added some new activation functions like sigmoid + tanh. They also work in the backpropagation process ofc.
A tiny micro autograd (automatic gradient) library that can do some calculations on scalar values. You can use it to implement some very basic neural networks.
Has no use at all because it's slow and impractical without any matrix operations. However, I learned something and had fun while building it! :)
Todo: Implement a neural network class that operates on Value objects so this becomes actually usable