Libra: Ergonomic Machine Learning

Test Image 1

Build Status Downloads Slack PyPi Release

Libra automates the end-to-end machine learning process in just one line of code. It is built for both non-technical users and software professionals of all kinds.

Full documentation can be found at


Install latest release version:

pip install -U libra

Install directory from github:

git clone
cd libra
pip install .

Alternatively you can build and use the docker image locally with:

docker build . -f docker/libra-normal/Dockerfile -t libra
docker run -v /path/to/my/data:/data -it --rm libra

Or if you have nvidia-docker installed.

docker build . -f docker/libra-gpu/Dockerfile -t libra-gpu
docker run -v /path/to/my/data:/data --gpus all -it --rm libra-gpu

Usage: the basics

The core functionality of libra works through the client object. A new client object should be created for every dataset that you want to produce results for. All information about the models that're built, the plots that are generated, and the metrics are created will be stored in the object.

You can then call different queries on that client object, and the dataset you passed to it will be used.

from libra import client

newClient = client('path/to/dataset') 
newClient.neural_network_query('please model the median number of households')

Now, calling

will return a dictionary of all the information that was generated:

dict_keys(['id', 'model', 'num_classes', 'plots', 'target', 'preprocesser', 
          'interpreter', 'test_data', 'losses', 'accuracy'])

Other queries can also be called on the same object, and will be appended to the models dictionary.

newClient.svm_query('predict the proximity to the ocean')

dict_keys(['regression_ANN', svm'])

Repo Not Found