/JetID-L1

Primary LanguageJupyter Notebook

Jet classification on FPGA

With synthesize.py, you can generate FPGA firmware for graph convolutional networks, interaction networks, and dense fully connected networks, for the models in https://github.com/sznajder/JetID-L1.

You need a Vivado certificate to run (e.g logging into geonosis from lxplus (ssh -XY geonosis)).

Everything is running Tensorflow 2.5, and a custom hls4ml branch.

Set up the Conda environment:

mamba env create -f environment.yml
mamba activate hls4ml-l1jets

To synthesize models:

python synthesize.py -C    # only create projects
python synthesize.py -C -T # create and trace model
python synthesize.py -C -B # create and synthesize projects
python synthesize.py       # Read reports