ZenDNN (Zen Deep Neural Network) Library accelerates deep learning inference applications on AMD CPUs. This library, which includes APIs for basic neural network building blocks optimized for AMD CPUs, targets deep learning application and framework developers with the goal of improving inference performance on AMD CPUs. 

ZenDNN v3.1 Highlights                     Downloads   User Guide

  • Enabled, tuned, and optimized for AMD 2nd and 3rd Gen EPYC processors 
  • Integrated with TensorFlow v2.5 
  • Features highly optimized primitives for AMD CPUs, including: Convolution, MatMul, Elementwise (Binary), BatchNorm, and ReLU 
  • Tested with a variety of neural network models across three major workload types:
    • Computer Vision: AlexNet, InceptionV3, InceptionV4, GoogLeNet, ResNet50, ResNet152, VGG16, VGG19
    • Natural Language Processing: BERT
    • Recommender Systems: DLRM, Wide & Deep
  • Supported on Ubuntu 18.04, Ubuntu 19.10, Ubuntu 20.04, RHEL 8.3, CentOS 8.3
  • INT8 models supported and tested with RESNET50 variants
ZenDNN is now open-sourced at https://github.com/amd/ZenDNN

Resources and Technical Support

Resources

Documentation

ZenDNN User Guide

TF + ZenDNN User Guide

Technical Support

To access 1×1 support, report issues, or request expert help, please email zendnnsupport@amd.com

Download

Please refer here for older versions of ZenDNN release: 

File Name Version Size Launch Date OS Bitness Description

File Name

Version

3.1

Size

292 MB

Launch Date

08/12/2021

OS

Ubuntu/ RHEL

Bitness

64-bit

Description

TF v2.5+ ZenDNN binary release package. MD5 Checksum: 6e8b6343083ceaa0f03807ce4021d141

File Name

Version

3.1

Size

135 MB

Launch Date

08/12/2021

OS

Ubuntu/RHEL

Bitness

64-bit

Description

TF v2.5+ ZenDNN binary release package for C++ API interface. MD5 Checksum: 3c518b9f87c455244649eb876bdd2173