MLflow Tracking - How to set Username, Source, Version and Model?

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

Hi Team,

I have been following the MLflow + Ray Tune integration documentation and examples, 1 2 3 + checking previous Github issues from both Ray and MLflow, and I can’t find a way to set Username, Source, Version and Model for the running experiment.

Any suggestions for how to set and capture these experiment meta data?

Thank you,
Heng

2

As an workaround for now, I am exposing custom tags as another parameters in MLflowLoggerCallback. Please let me know if this is a good approach. If so I can create a PR for this.

3

Hi @heng2j, this sounds like a good solution. Can you share what exactly you’re doing? I’m not very familiar with the Mlflow API, so knowing which functions you’re calling would be very helpful.

And yes, if you think this is missing functionality that other users would benefit from, a PR would be very much appreciated!

4

Thank you @kai , I have created a PR for this issue: 19417 adding custom tags in m lflow logger callback by heng2j · Pull Request #19532 · ray-project/ray · GitHub