It is known that Microsoft takes Machine Learning (ML) experiments from Data Scientists very seriously. A classic example of this is Azure ML having the capability to keep a 100% log of Machine Learning experiments.
Last year, Microsoft’s partner Databricks launched the MLflow project to handle similar tasks performed by Azure ML. The specialty of this project is that it is designed to work in almost any environment. This includes command lines, notebooks et al. resulting in a heavy spike in its popularity.
Both the companies are in close collaboration with Databricks’ integration into Azure natively supported by Microsoft. The collaboration is going one step further as Microsoft announced that it will be an active contributor for Databricks’ Machine Learning initiatives. The former has also extended native support to the project based in the Azure cloud.
Working with Machine Learning is no child’s play – tasks are highly complex and to an outsider, may even look unexplainable. However, the MLflow project aims to ease working with Machine Learning on a large scale. With MLflow, Data Scientists need not develop a model from scratch. They can just slightly tweak the pre-packaged code and get an array of results.
Data Scientists can also log –
- Miscellaneous files
- Deployable packaging of the ML model
The best part is, subsequently, when scripts are run, the results will log automatically – no manual intervention is required.
Facts about MLflow’s current status –
- APIs are available for Python and Java programming languages
- MLflow gets along really well with language-agnostic REST API
- The project gets downloaded 500,000 times a month
- 80 code contributors
- 40 contributing organizations
Now, since Microsoft will be contributing it will help standardize the entire technology stack for Databricks’ MLflow project including frameworks, AI DevOps, Cloud hosting et al.