Community

RUMUS is an open-source project built by SUM INNOVATION INC We welcome contributions, feedback, and collaboration from the Rust and deep learning communities.

Resources

GitHub Repository

Browse the source code, submit issues, and contribute to RUMUS.

View on GitHub

Issue Tracker

Report bugs, request features, and track development progress.

View Issues

Discussions

Ask questions, share ideas, and connect with other RUMUS users and contributors.

Join Discussions

Contributing

We appreciate all contributions — whether it's fixing a bug, adding a feature, improving documentation, or writing tests. Here's how to get started:

1

Fork the repository

Create your own fork of RUMUS on GitHub to start contributing.

2

Create a feature branch

Work on your changes in a dedicated branch with descriptive naming.

3

Write tests

Ensure your changes are covered by tests. Run cargo test to verify.

4

Submit a pull request

Open a PR against the main branch with a clear description of your changes.

Architectural Tenets

When contributing to RUMUS, keep these immutable design principles in mind:

The Tensor is Not a Junk Drawer

Storage, layout, and autograd are orthogonal — keep them separate.

Op-Driven Autograd

GradientStore is dumb. Operations own their reduction and broadcasting logic.

Borrow-Safe Optimizers

step(&mut grads) drains only registered ParamIds. No overlapping borrows, ever.

Zero-Allocation Inference

AutogradState::None + no_grad() RAII guarantees zero tape writes during inference.

Errors over Panics

User-triggerable faults return AutogradError. Panics are reserved for invariant violations.

Citation

If you use RUMUS in your research, please cite:

@software{rumus2025,
  title   = {RUMUS: A Native-Rust Deep Learning Framework},
  author  = {SUM INNOVATION INC},
  year    = {2025},
  url     = {https://github.com/SUM-INNOVATION/RUMUS},
  license = {MIT}
}