AI Tools / AI products & services
JAX is an open-source Python library for high-performance numerical computing and large-scale machine learning, developed by Google. It provides composable transformations of NumPy programs, including automatic differentiation (grad), just-in-time compilation (jit), and auto-vectorization (vmap), with execution on GPUs and TPUs via XLA.
A Python library for array-based computation that provides transformations for just-in-time compilation, automatic differentiation, and batching. The video explains how JAX runs computations on NVIDIA GPUs through XLA.
Trains models across multiple GPUs by declaring array placement and automatically inserting parallel work, including gradient averaging.
Builds and executes the model, including just-in-time compilation, ahead-of-time compilation, and portable artifact export for serving.