Burn is a next generation tensor library and Deep Learning Framework that doesn't compromise on flexibility, efficiency and portability.

Multi-platform high-performance compute language extension for Rust.

Models and examples built with Burn

Democratizing large model inference and training on any device.

Convert ONNX models into native, backend-agnostic Burn code for inference and fine-tuning.

CubeK: high-performance multi-platform kernels in CubeCL

Burn benchmarks

An 'xtask' CLI to manage standard repos and monorepos with easily reusable and extensible commands.

Rust system bindings for AMD ROCm HIP runtime used by CubeCL