This is the first release of MLflow TorchServe plugin (experimental). It is a new deployment plugin for MLflow that enables deploying PyTorch models built and trained in mlflow pipeline into TorchServe for running predictions.
Highlights
- Integrates TorchServe with MLflow Deployment plugin API
- Supports both command line (CLI) and python interfaces for managing deployments and running predictions
- Works with local or remote installation of TorchServe
- One can deploy both local models and models stored in the MLflow Model Registry