Published March 2, 2026 | Version v3
Dataset Open

Hyperoptimized Quantum Lego Contraction Schedules Supplementary Material

Authors/Creators

  • 1. ROR icon Duke University

Description

Hyperoptimized Quantum Lego Contraction Schedules

 

This repository incudes the code and data needed to reproduce all data and plots for the paper "Hyperoptimized Quantum Lego Contraction Schedules" by Balint Pato, June Vanlerberghe, and Kenneth R. Brown (2025).

 

Data and Images

All data and plots used in the paper are located in the `results` directory. Generate plots are in `results\images` and the data used to generate the plots is in `results/data`.

Setup Instructions

To set up the environment and install the required dependencies, follow these steps:
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

 

Running the Code

Contraction Cost Calculations & Tensor Sparsity

The script `contraction_cost_calculations.py` runs the Cotengra optimization and collects the intermediate tensor sparsities if specified. All command-line arguments are optional and have defaults if not included.
 
The command line arguments for this script are:
Argument Type Default Description
--file_name str "contraction_costs.csv" Name of csv file to save results
--num_runs int 100 Number of runs for each tensor network
--methods str (can be multiple) ["greedy", "kahypar"] Methods to run
--codes str (can be multiple) ["concatenated", "rotated", "rotated_msp", "rotated_tanner", "hamming_msp", "hamming_tanner", "holo", "bb_msp", "bb_tanner"] Tensor networks to run experiments for. Default is all available options.
--max_time int None Maximum runtime for Cotengra optimization
--max_repeats int 64 Maximum number of trials Cotengra will run
--sparsity_collection flag False When included, script will also collect sparsity information in a separate file

 

To gather the same data as shown in the paper, run the following command:
python src/contraction_cost_calculations.py --file_name "contraction_costs.csv" --num_runs 100 --max_repeats 64 --sparsity_collection

 

WEP Contraction Costs 

This script runs the Cotengra optimization and the WEP calculation. This data is used in the paper to compare the Cotengra dense cost to the custom SST cost. This script has the same command-line argument options and defaults as above.
 
To gather the same data as shown in the paper, run the following command:
python src/wep_calculations.py --file_name "wep_calculations.csv" --num_runs 100 --minimize "flops" --methods "greedy"

Optimal Costs

This script runs Cotengra's OptimalOptimizer for 4 small codes. Any larger or more complex codes were not feasible to compute optimally. The only command-line argument is the file_name.
python src/get_optimal_costs.py --file_name "optimal_costs.csv"

Generate Plots

To create the visualizations, run:
python src/plotting_functions.py
There are five plotting functions in this script, each corresponding to a plot from the paper. The current paths used are the pregenerated data in `results/data`. Make sure to update the script if wanting to use other data.

 

Files

hyper-optimized-contraction-schedules-for-qlego.zip

Files (88.6 MB)

Additional details

Software

Repository URL
https://github.com/junevanlerberghe/tn-contraction-optimization