Skip to content

Installation

System Requirements

  • Python ≥ 3.10 (tested with 3.11)
  • CUDA 12.1 (optional, for GPU acceleration)
  • 8GB+ RAM recommended

Step-by-Step Setup

1. Clone the Repository

git clone https://github.com/doctawho42/tgnn-solv.git
cd tgnn-solv

2. Create Conda Environment

conda create -n tgnn-solv python=3.11
conda activate tgnn-solv

3. Install PyTorch

Choose based on your system:

With GPU (CUDA 12.1): 

pip install torch --index-url https://download.pytorch.org/whl/cu121

CPU only: 

pip install torch

4. Install PyTorch Geometric

pip install torch-geometric -f https://data.pyg.org/whl/torch-2.4.0+cu121.html

5. Install TGNN-Solv

pip install -e ".[dev]"

This installs: - Core dependencies: torch-geometric, rdkit, numpy, pandas, requests, tqdm - Development tools: jupyter, matplotlib, scikit-learn, pytest, optuna

If you want the maintained Streamlit GUI as well:

pip install -e ".[gui,dev]"

This adds the dependencies used by:

  • scripts/launch_lab.py
  • scripts/gui/launch_lab.py
  • tools/experiment_lab/app.py

6. (Optional) Install External Baselines

For the maintained FastSolv / SolProp wrappers and benchmark helpers: 

pip install -e ".[baselines]"

This adds the dependencies used by:

  • scripts/external/run_fastsolv.py
  • scripts/external/run_solprop.py
  • scripts/experiments/run_external_baseline_benchmark.py
  • scripts/evaluation/benchmark_custom_model.py

If you want both the GUI and the external-baseline stack:

pip install -e ".[gui,baselines,dev]"

Verification

Verify installation:

import torch
import torch_geometric
from tgnn_solv import __version__

print(f"PyTorch: {torch.__version__}")
print(f"PyG: {torch_geometric.__version__}")
print(f"TGNN-Solv: {__version__}")

Docker Setup (Alternative)

If you prefer containerization:

docker build -t tgnn-solv .
docker run --gpus all -it tgnn-solv bash

The repository also ships a maintained compose layout for the current GUI and docs surfaces:

docker compose up lab
docker compose up docs

Available services include:

  • lab
  • Streamlit Experiment Lab on http://localhost:8501
  • docs
  • MkDocs site on http://localhost:8000
  • train
  • one tuned TGNN-Solv training run on the canonical split
  • evaluate
  • checkpoint evaluation
  • external-benchmarks
  • optional FastSolv / native SolProp benchmark runner

The Docker image now installs the grouped CLI layout, GUI extras, and docs toolchain by default. Optional external baselines stay behind a separate build switch because they are heavier and more environment-sensitive.

Troubleshooting

CUDA/GPU Issues

# Check CUDA availability
python -c "import torch; print(torch.cuda.is_available())"

# Reinstall with specific CUDA version
pip install torch --force-reinstall --index-url https://download.pytorch.org/whl/cu121

RDKit Import Errors

# RDKit sometimes needs conda installation
conda install -c conda-forge rdkit

Out of Memory

Reduce batch size in config files or use CPU-only mode for testing.

Next Step