Babygrad

A baby sized deep learning framework 👼🏻🔢

Usage

from babygrad.tensor import Tensor
from babygrad.visualize import graph

def f(x):
    return x**2 + x

def mse(y, y_hat):
    return (y-y_hat)**2

x = Tensor(9, requires_grad=True)
y = Tensor(100)

y_hat = f(x)

loss = mse(y, y_hat)
loss.backward()

graph(loss).render("out/test", view=True, format="png")