kaun ᚲ Documentation

Kaun is our PyTorch. It's the high-level deep learning framework built on top of Rune.

What kaun Does

Kaun gives you the building blocks for neural networks: layers, optimizers, training loops. If you've used PyTorch or Keras, you'll feel at home. Define your model, specify your loss, call train, kaun handles the rest.

The name comes from the rune ᚲ meaning "torch" or "fire." Fitting for a deep learning library.

Current Status

Kaun is in early development. The goal for alpha is training MNIST, a simple but complete workflow that proves the concept.

What's planned:

• Essential layers (dense, conv2d, dropout)
• Common optimizers (SGD, Adam)
• Training utilities
• Model serialization

This is enough to train real models. Everything else comes after we prove it works.

Design Philosophy

Kaun aims for PyTorch's flexibility, not Keras's high-level abstractions. You get building blocks, not black boxes. This means more code to write, but you understand exactly what's happening.

Learn More

• Getting Started - Build your first neural network
• Examples - (coming soon)