amrnet 0.1.1

Deep Learning library

AmrNet: Deep Learning Library

AmrNet is a lightweight deep learning library designed for simplicity and ease of use. It provides a set of basic tools to help you quickly build and train neural networks.

Installation

pip install amrnet==0.1.0

Implemented Features

• Tensors

• Layers :

  1. Linear

  2. ReLU

  3. Tanh

  4. LeakyReLU

  
  
  

• Neural Networks

• Loss Functions :

  1. MSE

  2. TSE

  3. MAE

  4. LogCosh

  
  
  

• Optimizers :

  1. SGD

  
  
  

• Data Utilities

• Training Utilities

Usage

Creating a Model

from amrnet.nn import NeuralNet

from amrnet.layers import Linear, Tanh,  ReLU



net = NeuralNet([

    Linear(input_size, hidden_size),

    Tanh(),

    Linear(hidden_size, output_size)

])

Training the Model

from amrnet.train import train



train(net, inputs, targets, num_epochs, data_iterator, loss, optimizer)

Predicting

predicted = net.forward(x)

Examples

Check out the examples directory for a variety of different projects using AmrNet.

License

MIT

TODO

• Add more layers

• Add more optimizers

• Add more loss functions

• Add more data utilities

• Add more training utilities

• Add more examples

• Add more tests

• Add more documentation