vgon 0.0.5

Variational Generative Optimization Network (VGON).

VGON Environment Setup

VGON scripts here require Python version 3.11 and above.

This guide provides instructions for setting up the VGON (Variational Generative Optimization Network) environment using either uv or pip.

Prerequisites

• Python 3.11 or higher
• Git

Method 1: Using uv (Recommended)

uv is a fast Python package manager and project manager, whose installation is simple as guide

Setup with uv

1. Clone the repository to local disk.

   git clone https://github.com/zhangjianjianzz/VGON.git && cd VGON
   

2. Create virtual environment if it wasn't contained

   uv venv .venv
   

   The python virtual environment will be located at .venv fold under root path in default. Linux/macOS users could activate it in shell by typing

   source .venv/bin/activate
   

3. Install dependecies in .venv for VGON scripts

   uv sync --extra full
   

   plot and notebook groups are optional for VGON itself.

4. Running experiments with uv

   # Run specific experiments
   uv run python BP/HXXZ/vgon_xxz.py
   uv run python Gap/Mix.py
   uv run python Degeneracy/H232/H232.py
   
   # Or activate environment first
   source .venv/bin/activate
   python BP/HXXZ/plot.py
   

Method 2: Using pip

Setup with pip

# Clone the repository
git clone <your-repo-url>
cd VGON

# Create virtual environment
python -m venv venv

# Activate virtual environment
source venv/bin/activate  # On Linux/macOS
# or
venv\Scripts\activate     # On Windows

# Upgrade pip
pip install --upgrade pip

# Install PyTorch (CPU version)
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

# For GPU support (CUDA 12.1), use instead:
# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

# Install the package with dependencies
pip install -e .

# Install optional dependencies
pip install -e ".[plot]"     # For plotting
pip install -e ".[dev]"      # For development
pip install -e ".[gpu]"      # For GPU acceleration
pip install -e ".[all]"      # Install everything

# Manual installation of ptitprince (if needed)
pip install git+https://github.com/pog87/PtitPrince.git

Verify Installation

Test your installation by running a simple example:

import torch
import pennylane as qml
import numpy as np

# Test basic functionality
print(f"PyTorch version: {torch.__version__}")
print(f"PennyLane version: {qml.__version__}")
print(f"CUDA available: {torch.cuda.is_available()}")

# Test quantum device
dev = qml.device("default.qubit", wires=2)
print("✓ PennyLane quantum device created successfully")

Project Structure

VGON/
├── BP/              # Barren Plateau experiments
│   ├── HXXZ/        # Heisenberg XXZ model
│   └── Z1Z2/        # Z1Z2 model
├── Degeneracy/      # Degeneracy detection experiments
│   ├── H232/        # H232 Hamiltonian
│   └── MG/          # Graph states
├── Gap/             # Nonlocality gap experiments
└── results/         # Experimental results

Running Experiments

Barren Plateau Experiments

# HXXZ model
python BP/HXXZ/vgon_xxz.py    # Run VGON training
python BP/HXXZ/vqe_xxz.py     # Run VQE baseline
python BP/HXXZ/plot.py        # Generate plots

# Z1Z2 model
python BP/Z1Z2/vgon_z1z2.py
python BP/Z1Z2/vqe_z1z2.py
python BP/Z1Z2/plot.py

Degeneracy Detection

# H232 Hamiltonian
python Degeneracy/H232/H232.py
python Degeneracy/H232/plot.py

# Graph states
python Degeneracy/MG/MG.py
python Degeneracy/MG/plot.py

Nonlocality Gap

python Gap/Mix.py              # Train gap model
matlab -batch "run('Gap/plotGap.m')"  # Generate MATLAB plots

Troubleshooting

Common Issues

1. CUDA not found: If you have a GPU but CUDA is not detected, reinstall PyTorch with CUDA support
2. PennyLane device errors: Ensure you have the correct PennyLane plugins installed
3. Memory issues: Reduce batch sizes in the configuration files
4. Import errors: Make sure all dependencies are installed correctly

GPU Setup

For GPU acceleration:

Development Setup

For contributors:

Support

• Paper: Commun Phys 8, 334 (2025)
• Issues: Create an issue on the GitHub repository
• Documentation: See individual module docstrings and comments

Development Containers

DevContainers' configuration in vscode/GitHub Codespace.