Google JAX
AI Development Tools

Google JAX

Google's Machine Learning Framework for Transforming Numerical Functions.


GoogleJAX is a machine learning framework for transforming numerical functions, Google calls it a combination of a modified version of Autograd (gradient function is automatically obtained by function differentiation) and TensorFlowXLA (Accelerated Linear Algebra). The framework is designed to follow the structure and workflow of NumPyas much as possible, and integrates with various Existing frameworks work together.
The main functions of JAX are to include:

  • grad: automatic differentiation
  • jit: compile
  • vmap: automatic vectorization
  • pmap: SPMD programming

Relevant Navigation

No comments

No comments...