Introduction

FOS is a Python framework that makes it easy to develop neural network models in PyTorch. Some of its main features include:

• Less boilerplate code required, see also the example below.
• Lightweight and no magic under the hood that might get in the way.
• You can extend Fos using common OO patterns.
• Get the insights you need into the performance of the model.

Installation

You can install FOS using pip:

pip install fos

Or alternatively from the source:

python setup.py install

Fos requires Python 3.6 or higher.

Usage

Training a model, requires just a few lines of code. First create the model, optimizer and loss function that you want to use, using normal PyTorch code:

model = resnet18()
optim = Adam(model.parameters())
loss = F.binary_cross_entropy_with_logits

Then create the FOS workout that will take care of the training and output:

workout = Workout(net, loss, optim)

And we are ready to start the training:

workout.fit(train_data, valid_data, epochs=5)

Examples

You can find several example Jupyter notebooks here

You can also run them on Google Colab directly:

• Basic https://colab.research.google.com/github/neurallayer/fos/blob/master/examples/basic_fos.ipynb
• MINST https://colab.research.google.com/github/neurallayer/fos/blob/master/examples/mnist_fos.ipynb
• Inputs https://colab.research.google.com/github/neurallayer/fos/blob/master/examples/inputs_fos.ipynb
• Tensorboard https://colab.research.google.com/github/neurallayer/fos/blob/master/examples/tensorboard_fos.ipynb

Contribution

If you want to help out, we appreciate all contributions. Please see the contribution guidelines for more information.

As always, PRs are welcome :)=