NEW! ZenDNN 1.5R is now available 

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 Processors, targets deep learning applications and framework developers with the goal of improving inference application performance on AMD CPUs.

ZenDNN v1.5R:

  • Target released for AMD EPYCTM processor family.
  • Contains ZenDNN Binary, TF wheel file (*.whl), End User License Agreement, ZenDNN User Guide
  • Integrated with TensorFlow 1.15
  • Features highly optimized primitives, including: Convolution, MatMul, Elementwise (Binary), BatchNorm, and ReLU for AMD CPUs
  • Tested with the following models: AlexNet, InceptionV3, InceptionV4, GoogLeNet, ResNet50, ResNet152, VGG16, VGG19, BERT, DLRM, Wide & Deep
  • Tested on Ubuntu 18.04, Ubuntu 19.10, Ubuntu 20.04, RHEL 8.2

Resources and Technical Support

Resources

Documentation

ZenDNN User Guide

Demo

The demo below compares SSDLite Mobilenet V2 inference performance using ZenDNN TF and vanilla TF on AMD EPYC 7742. For this object detection workload, about 53% improvement using ZenDNN TF over vanilla TF was observed.
 
Object Detection

Technical Support

ZenDNN is a well-supported library. Technical support is available to all customers via the channel below:

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

Download:

File Name Version Size Launch Date OS Bitness Description

File Name

Version

1.5

Size

95 MB

Launch Date

10/30/2020

OS

Ubuntu/RHEL

Bitness

64-bit

Description

ZenDNN binary release package. MD5 Checksum: 81a01ee3b12cd157f247821479b9f48b

File Name

Version

1.5

Size

5978 MB

Launch Date

10/30/2020

OS

Ubuntu/RHEL

Bitness

64-bit

Description

ZenDNN docker image release package. MD5 Checksum: a24a486ae1c02af96c02f8c192fbda01
Dependent Library AOCL-BLIS

File Name

Version

2.2.1

Size

6331 K

Launch Date

10/30/2020

OS

Ubuntu/RHEL

Bitness

64-bit

Description

AOCC compiled AOCL-BLIS library binary package. MD5 Checksum: 3cd3a60444335bec76ce63c91df755ed