fpgaConvNet Tutorial

A collection of tutorials for the fpgaConvNet framework. It is recommeneded to start with the end-to-end-example.

Setup

Python Tools

It is recommended to use conda to manage your python environment. Also, a python version above 3.8 is required.

conda create -n fpgaconvnet python=3.8

SAMO and fpgaConvNet packages are also required.

git submodule update --init --recursive
python -m pip install samo/ fpgaconvnet-model/ fpgaconvnet-hls/

Vivado

The fpgaconvnet-hls project relies on Vivado 2019.1. To install, first download from the Xilinx website.

Once installed, you will need to add a license server to your .bashrc file. You will also need to add Vivado to your path. To do so, add the following to your .basrc:

export PATH=/tools/Xilinx/Vivado/2019.1/bin:$PATH
export PATH=/tools/Xilinx/SDK/2019.1/bin:$PATH

You will need to setup JTAG drivers to program a device. To do so, execute the following script:

/tools/Xilinx/Vivado/2019.1/data/xicom/cable_drivers/lin64/install_script/install_drivers/install_drivers

For more information, visit here.

Bug Workarounds

The Y2K22 patch is required to fix issues with exporting designs in HLS. Please follow the instructions here.

Also, there is a known bug to do with C++ libraries. A workaround for this is adding the mpfr.h and gmp.h headers manually. For this project, you need to create a header file fpgaconvnet/hls/hardware/system.hpp which includes the following:

#ifndef SYSTEM_HPP_
#define SYSTEM_HPP_

#include "(path to Vivado 2019.1)/include/gmp.h"
#include "(path to Vivado 2019.1)/include/mpfr.h"

#endif

this isn't always necessary, so you can just create a blank header file if your system doesn't encounter this bug

Jupyter Setup

All the tutorials are done with jupyter notebook. Firstly, ipykernel and jupyter need to be installed.

conda install -c anaconda ipykernel jupyter

To run the notebooks locally, firstly you will need to add the python environment to ipykernel.

python -m ipykernel install --user --name=fpgaconvnet

Now you can start the notebook. Follow the instructions in the command line to view it in your browser.

jupyter notebook --no-browser --port=8888