A tiny scalar-valued autograd engine with a small PyTorch-like neural network library on top. Implements backpropagation (reverse-mode autodiff) over a dynamically built DAG. The DAG only operates over scalar values, so e.g. we chop up each neuron into all of its individual tiny adds and multiplies. However, this is enough to build up entire deep neural nets doing binary classification, as the demo notebook shows.
As it turns out, 90% of what you need for neural networks are a decent autograd/tensor library. Throw in an optimizer, a data loader, and some compute, and you have all you need.
What's next? Building a Tensor library?
Maybe.
Thanks karpathy/micrograd for your lessons :).
pip install tonygrad
Below there is a simple example showing a number of possible supported operations:
from tonygrad.engine import Value
"""tanh() VERSION"""
# inputs x1,x2
x1 = Value(2.0, label='x1')
x2 = Value(0.0, label='x2')
# weights w1,w2
w1 = Value(-3.0, label='w1')
w2 = Value(1.0, label='w2')
# bias of the neuron
b = Value(6.8813735870195432, label='b')
# x1*w1 + x2*w2 + b
x1w1 = x1*w1; x1w1.label = 'x1w1'
x2w2 = x2*w2; x2w2.label = 'x2w2'
x1w1x2w2 = x1w1 + x2w2; x1w1x2w2.label = 'x1*w1 + x2*w2'
#output
n = x1w1x2w2 + b; n.label = 'n'
#apply tanh to the output
o = n.tanh(); o.label = 'o'
#launch backprop on the built graph
o.backward()
For added convenience, the notebook trace_graph.py
produces graphviz visualizations. Here we draw the neural network graph built in the example above.
from tonygrad.trace_graph import draw_dot, trace
draw_dot(o)
The notebook demo.ipynb
provides a full demo of training an 2-layer neural network (MLP) binary classifier. This is achieved by initializing a neural net from tonygrad.nn
module, implementing a simple svm "max-margin" binary classification loss and using stochastic gradient descent for optimization. As shown in the notebook, using a 2-layer neural net with two 16-node hidden layers we achieve the following decision boundary on the moon dataset:
MIT