MedioGrad (WIP)

A ML library in the middle between Pytorch and Micrograd.

Not evoluted as TinyGrad since objective is to be numpy-based only (maybe in future with a bit of pycuda).

Stupid example

from tensor import Tensor 
import graph

a = Tensor([1,2,3])
b = Tensor([1,1,1])
c = a + b
c.backward()
graph.draw_dot(c, format="png").render(filename="graph")