How does MLflow compare to ModelDB (A system to manage machine learning models)?

[Emre Sevinç]

Hello,

I wonder how MLflow compares to ModelDB (A system to manage machine learning models).

In ModelDB web page, I've seen Matei Zaharia as one of the contributors:

    https://mitdbg.github.io/modeldb/

Any ideas?

--

Emre

[Matei Zaharia]

Hi Emre,

I did indeed work on the ModelDB research project back in 2016. MLflow Tracking is the closest part to ModelDB in that it lets you report experiment results and parameters, so it solves a similar problem. There are some differences in how these systems work though — for example, right now, MLflow Tracking is a lower level API you call directly, whereas ModelDB adds wrappers around popular ML libraries (SciKit-Learn and Spark MLlib) to easily capture this info. The ways they store and display results are different too.

Matei

> --
> You received this message because you are subscribed to the Google Groups "mlflow-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to mlflow-users...@googlegroups.com.
> To post to this group, send email to mlflow...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/mlflow-users/1468ddd1-03a1-47dd-9844-af8b5fdfb28b%40googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.