In this folder we explain how to deploy Concrete ML models. We show-case how to do this on 3 examples:
- Breast cancer classification using a simple XGBoost model
- Sentiment analysis by running a XGBoost model on top of a Transformer model
- CIFAR-10 classification using a VGG model split in two parts.
You can run these example locally using Docker, or on AWS if you have your credentials set up.
For all of them the workflow is the same: 0. Optional: Train the model
- Compile the model to a FHE circuit
- Deploy to AWS, Docker or localhost
- Run the inference using the client (locally or in Docker)
The script to deploy the model compiled to a FHE circuit is the same for all. The main difference between them is the client. Each use-case needs its own client.