Praxis is a core high-performance machine learning library used within the Pax framework. It provides layer, model, configuration, input, and optimization components for large-scale neural network training.
A core, high-performance ML library of Pax. Praxis is often called the "Layer library". Praxis contains not just the definitions for the Layer class, but most of its supporting components as well, including: - data inputs - configuration libraries (HParam and Fiddle) - optimizers Praxis provides the definitions for the Model class.