sop-heuristic-opt

Stochastic orienteering problem self-play pipeline leveraging GNNs with MCTS

TODO List:

• Add support for torch dataloader
• MILP script, maybe add multithreading
• Proper experiment pipeline
• MCTS ACO Implementation

Installation

1. Create a new environment (ex. w/ conda)

conda create -n <ENV_NAME> python=3.10
conda activate <ENV_NAME>

2. Install packages in requirements.txt

pip install -r requirements.txt

3. Install Pytorch and PyG w/ torch_scatter & torch_sparse from links

Using Docker

Install

Using Linux (Ubuntu) install nvidia-toolkit

sudo apt install ubuntu-drivers-common
sudo ubuntu-drivers devices
sudo apt install nvidia-driver-XXX
sudo reboot now

Where XXX is based on recommended version listed in second command.

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit nvidia-container-toolkit
sudo reboot now

Add the following to your rc file of your shell of choice:

export PATH=/usr/local/cuda/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-12.<version>/lib64\
                         ${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}

Where <version> is the version you find in /usr/local/.

After you've added this, make sure to source your rc file.

Running Container

make build
make bash

Execution

This will run improve_heuristic.py.

make run

To generate a dataset

make dataset

You can find the outputs in data/.

Source Control

To use precommit checks, make sure that you install pre-commit and initialize your repo. If you're using Linux you can make repo-init and it will install all hooks into your local .git folder. If not, install pre-commit and then pre-commit install to initialize the repo. For example on Mac, brew install pre-commit will do the trick.