Abstract for IBM Z Deep Neural Network Library Programming Guide and Reference
IBM Z® Deep Neural Network Library (zDNN) provides high-level libraries that enable frameworks and model compilers to use the IBM Z on-chip AI accelerator, known as IBM Z Integrated Accelerator for AI.
zDNN provides the following benefits to simplify access to the IBM Z Integrated Accelerator for
AI:
- A high-level language application programming interface (API) for supported IBM Z Integrated Accelerator for AI primitives
- A unified approach for tensor data layout and type conversion
- Change management and version control mechanisms for IBM Z Integrated Accelerator for AI feature-level compatibility
This document contains information and reference material about the zDNN API, which is a standard C library.
This document also includes information and reference material about the
IBM Z Artificial Intelligence
Optimization Library and the
IBM Z Artificial Intelligence Data
Embedding Library.


The IBM Z Artificial Intelligence Optimization Library provides a set of APIs that offer a seamless interface to various methods of hardware acceleration for programs that use semantic-based mathematical operations designed around vector and matrix operations.
- The IBM Z Artificial Intelligence Data Embedding Library provides a set of APIs to access a collection of packages designed to build and process vector embedding models on z/OS.
Who should read this information
This information is intended for use by application developers who wish to use the zDNN, IBM Z Artificial Intelligence
Optimization Library,
and IBM Z Artificial Intelligence Data
Embedding Library APIs to create AI applications.
Related information
For additional information about z/OS®, see z/OS Information Roadmap.