Sonnet is a library built on top of TensorFlow 2 that provides simple, composable abstractions for machine learning research. It centers around a single concept, snt.Module, which can hold parameters, other modules, and methods that apply functions to user input. The library ships with predefined modules like linear layers, convolutions, and batch normalization, along with predefined networks such as MLPs. Unlike many frameworks, Sonnet is unopinionated about training and does not ship with a training framework, allowing users to build or adopt their own.