MLflow is a platform to streamline machine learning development including:
- tracking experiments,
- packaging code intro reproducible runs
- sharing and deploying models
MLflow is a set of lightweight APIs which can be used with any ML application or library (TensorFlow, Pytorch, XGBoost) in any environment used to run ML code (notebooks, standalone applications, or the cloud).
MLflow components
- Tracking:
- Log parameters, code, results from ML experiments and compare them using an interactive UI
- Projects:
- Code packaging format for reproducible runs using Conda and Docker
- Models:
- Model packaging format and tools allowing easy deployment of the same model to batch and real-time scoring on platforms like Docker, Apache Spark, Azure ML, and AWS Sagemaker
- Model Registry: Centralized model store, set of APIs, and UI to collaboratively manage full lifecycle of MLflow Models