This project provides the example inference code of generating optimizers and controlling their hyper-parameters for the 20 problem instances in the jupyter notebook.
Firstly, create the conda environment with python 3.9.18 and torch 2.3.1:
conda create -n DesignX python=3.9.18
conda activate DesignX
conda install pytorch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 pytorch-cuda=12.1 -c pytorch -c nvidia
pip install -r requirements.txtThen after selecting the jupyter notebook kernel as DesignX, we can execute the inference code step by step and see how DesignX solves the problems.
Besides, the complete results of each baseline across all 3200 synthetic testing problem instances mentioned in Section 4.1 are in the excel table.