A simple autograd engine along with a PyTorch/Tensorflow like API module to easily train neural nets
Project description
Micronets
Micronets is a python framework for the simple implementation and training of deep neural networks. The project consists of a small auto-gradient engine module which implements back propagation through a DAG (Directed Acyclic Graph) and a module resembling the PyTorch/Tensorflow APIs to create and train dense neural networks on top of it.
The package also provides additional functionalities such as plotting of the training curve, make predictions and visualisation of computational graphs after training
Installation
pip install micronets
Usage
1. Building Computational Graphs
from micronets.autograd_engine import Value
from micronets.graph import draw
#inputs
x1 = Value(2.0,label='x1')
x2 = Value(4.0, label='x2')
#weights
w1 = Value(-3.0,label='w1')
w2 = Value(1.0,label='w2')
#bias
b = Value(3.3,label='b')
x1w1=x1*w1; x1w1.label='x1*w1'
x2w2=x2*w2; x2w2.label='x2*w2'
x1w1x2w2 = x1w1+x2w2; x1w1x2w2.label='x1w1 + x2w2'
z = x1w1x2w2 + b; z.label='z'
output = z.tanh(); output.label='output'
output.backward()
draw(output)
2. Neural Net Implementation
from micronets.nn import DNN # Dense Neural Network
from micronets.losses import binary_cross_entropy
xs = [
[2.0, 3.0, -1.0],
[3.0, -1.0, 0.5],
[-0.5, -1.0, -1.0],
[-1.0, -1.0, 0.5]
]
ys = [1.0, 0.0, 1.0, 0.0]
model = DNN(input_features=3, layers=[2,2,1], activations=['relu', 'relu', 'sigmoid']) # creates model architecture
history, graph = model.train(X=xs, Y=ys, iterations=500, loss_function=binary_cross_entropy, learning_rate=0.1) # trains network and stores loss function history and computational graph
predictions = model.predict([ # outputs predictions
[2.0, -1.0, 3.5],
[-1.0, -3.0, 0.5]
])
model.plot(history) # plots learning curve
Additionally the playground.ipynb
file demonstrates functionality through a more complex example along with its outputs.
Lastly, a big thank you to Andrej Karpathy for inspiring this project :)
License
MIT
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
File details
Details for the file micronets-0.1.2.tar.gz
.
File metadata
- Download URL: micronets-0.1.2.tar.gz
- Upload date:
- Size: 5.9 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.6
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 0c500f912492789f1e22c0719e4cab1168c765c8419d538d33a7fe1a23e0762e |
|
MD5 | 1657931f5607e447f54f3aaa879abc3a |
|
BLAKE2b-256 | bb0bba0b3fc09bbefa3deefe5d198c2f9c65c2d73739c91e0f3fd92627de26f8 |
File details
Details for the file micronets-0.1.2-py3-none-any.whl
.
File metadata
- Download URL: micronets-0.1.2-py3-none-any.whl
- Upload date:
- Size: 6.6 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.9.6
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 4b28b43c0be12f308f92e60a07af84e16fce5d34746eed29cd8c015ce78a853e |
|
MD5 | c2c3b22169a9e97f869f19182642b5cc |
|
BLAKE2b-256 | d808ed92c87bd321d3647991040445e87e936ecfcc335e2668e3275fe700b3ed |