TorchDrug Use TorchDrug as a modular PyTorch graph-learning stack: 1. load a dataset, 2. choose a representation model, 3. wrap it in a objective, 4. train and evaluate it with . The current official documentation and latest published release are both 0.2.1 (released July 2023; rechecked October 1, 2026). Treat newer Python or PyTorch combinations as unverified rather than silently assuming compatibility. Start with the version guard Before generating or debugging code, inspect the environment: The supported matrix for TorchDrug 0.2.1 is: - Python 3.7 through 3.10 - PyTorch 1.8 through 2.0 -…