FATE-Serving is a high-performance, industrialized serving system for federated learning models, designed for production environments. for more detail, You can click WIKI for more information for more information
FATE-Serving now supports
- High performance online Federated Learning algorithms.
- Federated Learning online inference pipeline.
- Dynamic loading federated learning models.
- Can serve multiple models, or multiple versions of the same model.
- Real-time inference using federated learning models.
- Support multi-level cache for remote party federated inference result.
- Support pre-processing, post-processing and data-access adapters for the production deployment.
- Provide service managerment for grpc interface by using zookeeper as registry
- Requests for publishing models are persisted to local files，so the loaded model will be loaded automatically when the application is restarted