MLflow — Open source Machine Learning platform
MLflow is an automated machine learning platform that aims to unify ML and AI workflows. It is designed to work like a charm with any ML library, algorithm, language and deployment tool.
One of the most strenuous tasks involved in data science development process is to get the model into production. To overcome this humongous obstacle, the creators of Apache Spark have come up with an unified solution to all machine learning framework challenges — MLflow.
In MLflow library, the Machine Learning lifecycle is transparent and standard. It also helps in tuning and makes the deployment phase a cakewalk. MLflow is an open source machine learning platform that manages the entire ML lifecycle from start to production.
Remarkable advantages of MLflow are :
- User friendly.
- One of the add-on feature is that the ML model can be served with REST API.
- Easy integration with any cloud, python and a number of open-source machine learning frameworks including Apache Spark, TensorFlow, and SciKit-Learn.
- Effortless comparisons between various models.
- Highly scalable experiments.
- Centralized environment to log the model’s training values.
The current version of MLflow has three components packed along with it, they are :
- Tracking: For querying about the performance of the model and recording data on experiments
- Projects: Provides a simple format for reproducing code and models.
- Models: Smooth management and deployment of machine learning models into production.
Installation process of MLflow right now is quite straight forward using pip :
pip install mlflowSome of the alternatives to MLflow are Sagemaker, Sacred and FGLab.
We would highly recommend researchers who want to compare the models and machine learning enthusiasts who look to serve machine learning models through REST APIs to checkout MLflow.
Thanks for reading.
Reach out to me on Linkedin at
