Keras
High-level modelling API that runs on several backends.
Overview
Keras describes a model declaratively and runs it on whichever backend is configured, so the same definition works on TensorFlow, JAX or PyTorch. It solves the problem that a model's structure and its execution engine are separate concerns, and most people only want to express the first. It exists because that separation of concerns is what makes a model definition readable and portable, and it has become the recommended high-level API for TensorFlow.
Technology Graph
How Keras connects to the rest of the stack.
Tools
Software that implements or integrates with Keras.