Compiler and AI framework developer resources

Developing for the Integrated Accelerator for AI

Most users of the IBM z16 Integrated Accelerator for AI will be accessing it through frameworks and technologies like the IBM Deep Learning Compiler, TensorFlow, or Snap ML. The other pages on this site are focused on this type of usage. Information on using these can be found throughout these pages.

For those that wish to develop or enhance AI frameworks and compilers to leverage the Integrated Accelerator for AI, a lower level SDK is needed. To facilitate this, IBM has created the zDNN library. IBM zDNN is the IBM Z Deep Neural Network library, which features C apis that can simplify use of the z16 accelerator.

Details and examples can be found in the IBM zDNN github.

Note that zDNN library is distributed in z/OS 2.4 and above, as well as on RHEL, Ubuntu, SUSE at various levels.