This repository contains a pytorch implementation of AdaSTE: An adaptive Straight-Through Estimator to Train Binary Neural Networks The code is built on top of the implementation for "Training Binary Neural Networks using the Bayesian Learning Rule". https://arxiv.org/abs/2002.10778
To reproduce the AdaSTE experiments run the shell scripts located in /scripts.