Question: what framework/language does Designer use behind the scenes?

Question

I'm using Azure ML Designer, but it isn't clear what framework/language is being used behind the scenes. For example, deep learning is incredibly computation-heavy, so I would like to run any matrix multiplications on a GPU instead of on a CPU. For the builtin modules, does the framework support GPU operations, or only CPU? If I have a complex pipeline, are all steps in the pipeline (that don't depend on each other) run in parallel? Or does that require the use of a Compute Cluster instead of a Compute Instance? How are NaNs handled in the dataset? So far I've only encountered seg faults with the builtin models, so I'm assuming they aren't handled. Also, why does the pipeline take so long to run? If I code things in pure pandas, the pipeline should finish in under a second on my laptop, but takes several minutes to run in Azure ML Designer, regardless of which hardware I choose.

Answer 1

@Adam Stewart Thanks for the questions. For any errors that are internal or faults our PG team looks at the issues that are submitted from Azure ML portal. The smiley option on the top right corner of the portal helps you to send these errors and queries related to the pipelines or experiments to our service engineers who can lookup the issue for more details and advise.

It would be great if you could also email us at AzCommunity[at]microsoft[dot]com to discuss more about the feature requests and questions as some of them might be in development or some may need more information to create a backlog.