Keras model and weights for GarNet-on-FPGA
Y. Iiyama, Jan Kieseler · DSpace@MIT (Massachusetts Institute of Technology) · 2020
Source code and input weights data for GarNet-on-FPGA This repository contains the Keras layer and model files for the simultaneous regression and classification task described in arXiv:2008.03601. Dataset in 10.5281/zenodo.3888910 can be fed into the training script (train.py) to reproduce the results in the paper. The GarNet layer file (garnet.py) is a partial copy from the original at commit 6d1127d. The hls4ml configuration and weights files for the "continuous" and "quantized" models in the paper (Vmax = 128) are also packaged together in the tarball hls4ml_inputs.tar.gz. HLS workspaces can be created from the input files with hls4ml version > v0.3.0 (commit e86c103 and later)