Primus

Primus is a generic distributed scheduling framework for machine learning applications, which manages training lifecycle and data distribution for machine learning trainers such as TensorFlow to perform distributed training in massive scales. For more information, please visit overview and quickstart.

Building Primus

Primus is built using Apache Maven. To build Primus and its example programs, run:

mvn -DskipTests clean package

For general deployments, environment configurations are required, see the setups in quickstart for references.

Contribution

Primus is under active development, and we use GitHub issues for tracking requests and bugs, feel free to contact us for any assistance. If you want to contribute to Primus, you are expected to uphold our code of conduct.

Security

If any potential security issue is discovered, please do not create a public GitHub issue. Instead, do inform Bytedance Security directly via our security center or vulnerability reporting email.

License

Apache License 2.0