Cerbero

Multi-task learning made easy

Cerbero aims to pick up the multi-task learning framework that was started by Snorkel before they dropped support. Rather than focusing on weak supervision and labeling functions Cerbero will mainly focus on allowing users to build and train multi-task models with minimal fuss.

Quickstart

Cerbero requires Python 3.6 or later. To install Cerbero, clone the environment and pip install

git clone git@github.com:strongio/cerbero.git
cd cerbero
pip install -e .

Examples

There are a couple of examples which go over basic usage of Cerbero and the multi-task framework:

• Basic Example: simple example showcasing multi-task learning with Cerbero on synthetic data
• CIFAR-10 Example: A classification/regression multi-task example using the CIFAR-10 dataset

Development Environment

Following it's predecessor Snorkel, Cerbero uses tox to manage development environments To get started, install tox, clone Cerbero, then use tox to create a development environment.

git clone git@github.com:strongio/cerbero.git
pip3 install -U tox
cd cerbero
tox --devenv .env

Running tox --devenv .env will install create a virtual environment with Cerbero and all of its dependencies installed in the directory .env. This can be used in a number of ways, e.g. with source .env/bin/activate.

Testing

At the moment there are just a few tests which we can run using the following tox commands:

tox -e py37  # Run unit tests pytest in Python 3.7
tox -e check # Check style with black
tox -e type # Run static type checking with mypy
tox -e fix # Fix style issues with black