In this repository, we serve several example trained models to make predictions over HTTP.
- Sentiment Classifier (PyTorch)
- MNIST Digit Classifier (PyTorch Lightning)
- Fine-Tuned Language Model for Sentiment Classification (HuggingFace)
- Fine-Tuned Language Model for Question Answering (HuggingFace)
- Fine-Tuned Language Model for Summarization (HuggingFace)
Install catacomb by running:
pip install catacomb-ai
For each example, run catacomb in the example directory to deploy the machine learning instance to Catacomb.
In each example directory, custom machine learning system code is defined in system.py, where we only load a pre-trained model (i.e. no training occurs within this application). In particular, we implement the output() interface on Catacomb's System class, which is called in our generated server.py file and Dockerfile.
MIT