Frequently while training Machine Learning models on Azure, the loss and accuracy output freezes half way through training. A "canceled" message in red is appearing the notebook cell. However, the kernel is still displaying a "busy" message.
I am paying for a GPU and this is making the service unusable for me as I don't know if the model is training correctly.
It appears that I have to pay for technical support to register a ticket. What a joke!!!