What is Squirrel?

Squirrel is a Python library that enables ML teams to share, load, and transform data in a collaborative, flexible, and efficient way.

1. SPEED: Avoid data stall, i.e. the expensive GPU will not be idle while waiting for the data.

2. COSTS: First, avoid GPU stalling, and second allow to shard & cluster your data and store & load it in bundles, decreasing the cost for your data bucket cloud storage.

3. FLEXIBILITY: Work with a flexible standard data scheme which is adaptable to any setting, including multimodal data.

4. COLLABORATION: Make it easier to share data & code between teams and projects in a self-service model.

Stream data from anywhere to your machine learning model as easy as:

it = (Catalog.from_plugins()["imagenet"].get_driver()
      .get_iter("train")
      .map(lambda r: (augment(r["image"]), r["label"]))
      .batched(100))

Check out our full getting started tutorial notebook. If you have any questions or would like to contribute, join our Slack community.

Installation

You can install squirrel-core by

pip install squirrel-core

To install all features and functionalities:

pip install "squirrel-core[all]"

Or select the dependencies you need:

pip install "squirrel-core[gcs,torch]"

Please refer to the installation section of thedocumentation for a complete list of supported dependencies.

Documentation

Read our documentation at ReadTheDocs

Example Notebooks

Check out the Squirrel-datasets repository for open source and community-contributed tutorial and example notebooks of using Squirrel.

Contributing

Squirrel is open source and community contributions are welcome!

Check out the contribution guide to learn how to get involved.

The humans behind Squirrel

We are Merantix Momentum, a team of ~30 machine learning engineers, developing machine learning solutions for industry and research. Each project comes with its own challenges, data types and learnings, but one issue we always faced was scalable data loading, transforming and sharing. We were looking for a solution that would allow us to load the data in a fast and cost-efficient way, while keeping the flexibility to work with any possible dataset and integrate with any API. That's why we build Squirrel – and we hope you'll find it as useful as we do! By the way, we are hiring!

Citation

If you use Squirrel in your research, please cite it using:

@article{2022squirrelcore,
  title={Squirrel: A Python library that enables ML teams to share, load, and transform data in a collaborative, flexible, and efficient way.},
  author={Squirrel Developer Team},
  journal={GitHub. Note: https://github.com/merantix-momentum/squirrel-core},
  doi={10.5281/zenodo.6418280},
  year={2022}
}