Let me introduce you mlFlow | Part I

Earlier it was quite difficult to track things down, data scientists were not able to track the progress during the model development and tuning. ML flow is a tool which performs all such functions. Lets get started …

In this short article i will just introduce mlFlow and its four components. If you want to add one new gem in your skills, follow this series for upcoming articles.

ML flow is a very effective and open source platform to manage the ML lifecycle, that includes, experimentation, deployment, reproducibility long with a central model registry. In other words, it is an easy to use and open source platform for the complete machine learning lifecycle.

ML flow has it’s four components:

1. ML flow Tracking: It is an important component that keeps a track of runs by saving parameters, tags and metrices. It allows us compare and visualize them in the browser, in any way we want.
2. ML flow Model: Its is the standard format for packaging the machine learning models. It is really helpful in variety of downstream tools, like, Apache Spark. It enables us to save a model in different functions (python function, R function).
3. ML flow Project: It is a really useful format which allows us to package the data science code in a reproducible and reusable way. Moreover, it includes command-line tools and an API to chain together the projects into particular workflows.
4. ML flow Registry: Its is a centralized model store which has set of APIs and UI to manage the full lifecycle of an ML flow Model. It helps in model versioning, stage transitions, annotations and model lineage.

ML flow enables us to step forward towards reproducible data science and model deployment. Very soon i will be coming with a new article which will explain how to get started with mlFLow.

Thats all for the day…