PyTorch Hub Model Weights Cache Folder (~/.cache/torch)

Bottom Line: It is 100% safe to delete the PyTorch model weights cache. Deleting ~/.cache/torch will NOT delete your Python AI scripts, Jupyter Notebooks, or custom training datasets.

What is the PyTorch Cache Folder?

Located at ~/.cache/torch on Linux/macOS (and %USERPROFILE%\.cache\torch on Windows), this folder is created by PyTorch and PyTorch Hub.

When running deep learning models (ResNet, YOLO, Whisper, Stable Diffusion, or vision transformers), PyTorch downloads pre-trained model weights (.pt, .pth, .bin files) into checkpoints/ and hub/. Because AI model weights weigh 500 MB to 5 GB each, experimenting with multiple neural networks hoards 15 GB to 50 GB+.

Can You Delete PyTorch Cache?

✅ Yes, it is completely safe to delete.

• No code loss: Your Python training code and model architectures remain 100% safe.
• Auto-downloaded: PyTorch will simply re-download pre-trained weights from PyTorch Hub when you run model inference scripts.

How to Safely Clean Up PyTorch Model Cache

Method 1: Delete via Terminal

# macOS / Linux
rm -rf ~/.cache/torch/hub/checkpoints/*

# Windows (PowerShell)
Remove-Item -Recurse -Force $env:USERPROFILE\.cache\torch\hub\checkpoints\*

Related Guides

• Learn about deleting Hugging Face cache.
• How to clean up pip cache.
• Check if you can clean Python venv folders.