10-09-2019 11:52 AM
Yes you can! Follow the DNNDK User Guide:
CPU version DECENT can perform quantization on CPU-only host machine without GPU installed; however, this requires longer running time.DNNDK provides a unified solution for deep neural network inference applications by providing pruning, quantization, compilation, optimization, and run time support. The Deep Compression Tool, DECENTTM, employs coarse-grained pruning, trained quantization and weight sharing to address these issues while achieving high performance and high energy efficiency with very small accuracy degradation. Here is a sample DECENT Quantization script. https://www.xilinx.com/support/documentation/user_guides/ug1327-dnndk-user-guide-190201.pdf#page=36
You can scroll down in the link to see more and the workflow.
This page has the method for DECENT: https://www.xilinx.com/support/documentation/user_guides/ug1327-dnndk-user-guide-190201.pdf#page=45
Hope that helps!
11-03-2019 05:17 PM
I know how to quantify according to the method in the manual, but I want to use the quantization method outside the manual. Is this ok?
11-04-2019 11:39 PM - edited 11-04-2019 11:40 PM
No. For now a quantization result from third party tools is not officially supported by DNNDK.
The interest question is that:
Why would you like to use a third party quantization tools to quantize a pb model?
That may help our developer team to improve their tools. :-)
11-05-2019 01:24 AM
Me too. Please feel free to post or ping me if you find anything interesting through your investigation. :-)