Is there a way to export experiment parameters and logged metrics in Azure ML to CSV?

Question

I am running a bunch of ML experiments using AzureML, sometimes changing training parameters and sometimes aspects of the data preprocessing. In general, for a given experiment I will be able to get a table (aka "view") like this:

While the UI allows some minimum level of customization, sorting runs by e.g. desired columns (say the accuracy to identify the best runs) seems really problematic.

The only workaround I am aware of is to save the page to HTML (!) and extract the values from there.
The data in the cells can't by copied with a cursor either...

Is there an easy way to export the data collected during several runs, via the UI or programmatically, without the need to scrape the blob storage of the Azure ML workspace (I am asking the community here as docs don't seem particularly helpful)?

Answer 1

@Davide Fiocco You can use the SDK to export the run/s using Run() or get_runs()
This sample provides an example to get all the files associated with a run. I think you can use the SDK to get all the runs with get_runs() and load the list to a dataframe and export the same to csv.

The portal does not have the functionality to export the table but we will provide the feedback to our team for upcoming updates to the portal. Thanks!!