We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

Showing results for 
Search instead for 
Did you mean: 

DNNDK v3.0 Has Been Released!

Xilinx Employee
Xilinx Employee
0 0 1,873

DNNDK v3.0 was just released yesterday (Apr 30th). This is a major upgrade from DNNDK v2.08 which was released in February. The download is available on Xilinx.com.

The key features are as follows:

  • Tensorflow support – Matching all the functionalities previously available for Caffe, including model quantization, optimization, compilation, run-time libraries and profiling will be available directly with TensorFlow. Users who are familiar with the Caffe flow should be able to get up and running with TensorFlow very quickly. Popular models such as Resnet, Googlenet, SSD, Mobilenet and Yolo are supported.

  • Along with the toolchain, a brand-new AI SDK is also included in this release. It offers a set of high-level application-oriented APIs and optimized libraries in addition to the DNNDK runtime. These new features allow users with zero FPGA or AI/ML knowledge to be able to develop efficient AI/ML applications with Xilinx hardware acceleration.

  • All the compilation / building is performed on the host machine, unlike with v2.08, which generated the final executables on the target board. You no longer need to copy / transfer the DPU kernel files from host to target. This makes the whole process quicker and easier.

  • This release includes a newer version of the DPU (v1.4.0) which now supports all Xilinx DNN models. 

There are some other minor updates in DNNDK v3.0, please see the user guide for details.

Tags (1)