Shurl is Shrunk's artificial intelligence platform. This is not meant to be a standalone project and is supposed to be a submodule for Shrunk, an open-source URL shortener for Rutgers University.
Note
This research project is not officially endorsed by Rutgers University.
- Produces title suggestions for the shortened URL
- Interactive web app for demonstration purposes
We use pre-commit to ensure that all committed code does not violate the Ruff linter.
- Fork, then clone the repository
- Install pre-commit
pip install pre-commit --break-system-packages && pre-commit install
- Launch the service.
docker-compose up
When training the transformer model with the Shrunk database, each iteration took a significant amount of time (approximately 60 seconds per iteration on an Apple M2 Max with 32GB of RAM). Consequently, it is advisable to train the model in the cloud rather than on-device.
Instructions for training the model on Amazon Web Services coming soon.
- Shrunk, https://github.com/oss/shrunk
- Shurl Trainer (for SageMaker), https://github.com/novialriptide/shurl-sagemaker-trainer
- Hugging Face Inc, Transformers: State-of-the-Art Natural Language Processing
- Google Research, Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
- Google Research, LongT5: Efficient Text-To-Text Transformer for Long Sequences