Flux is a 100% pure-Julia machine learning stack offering lightweight abstractions on top of Julia's native GPU and automatic differentiation (AD) support. Designed for researchers, scientists, and developers seeking high performance and flexibility, Flux allows you to define and train models in an intuitive, mathematical style with compiled, GPU-optimized code. Its extensible ecosystem includes packages for computer vision, graph neural networks, probabilistic programming, and more, making complex architectures and scientific ML development accessible and customizable.
Visit Flux's official website for product details and getting started.