Azure Machine Learning Studio training 24 hour timeout

Question

I'm a newbie trying out Azure's Machine Learning (ML) Studio module. I own a standard subscription level account which grants me an experimental duration of "Up to 7 days per experiment with a maximum of 24 hours per module" according to the ML Studio's pricing site.

However, since my dataset is extremely large, I would need a much longer training duration than the allocated 24 hours (I have tried and it timeout-ed even with the simplest NN architecture). Is there workaround for this issue?

Thank you in advance.

Answer 1

I would suggest to stop using Azure Machine Learning Studio and switch to "real" Azure ML with Azure Machine Learning Services, where you will have much more control on your compute needs.

Azure ML Studio roadmap is really limited and the purpose of this solution was to help people coming to Machine Learning. If you have a real use-case, use Azure Machine Learning Services.

Answer 2

Machine Learning Studio Workspace, which offers users a serverless, drag-n-drop environment and powered by CPU only, and there is no way for you to add GPU resource. It is the reason that your training for more than 24 hours.

The only way is to switch to Machine Learning Service where you could connect to GPU via Azure notebook to boost your training.