Tips

Logging

Instead of using run.log, you may also use mlop.log to log metrics. However, note that this behavior implicitly defines run to be the latest started process in the mlop.ops group and therefore may not be what you are looking for.

To retrieve the active runs, you can call mlop.ops which returns an array of all the runs.

To dive into the details of a run, check out the experiment section.

Grouping

The logger additionally supports automatic grouping by using a / in the log name.

run = mlop.init(project="mlop", name="experiment")
 
group = "val"
run.log(f"{group}/name", value)

This will group all the logs under the val (validation) group and will be displayed in the UI as a separate entity.