Reputation: 48436
Training a simple convolutional network to recognize MNIST digits on Microsoft Azure (in Machine Learning Studio) takes many many times longer than it does for (already very slow) learning of exactly the same model locally, on a CPU (MacBook Pro, with limited memory) with TensorFlow.
Is there a way — perhaps purchasing resources or connecting virtual GPUs — to improve performance of Azure Machine Learning?
Upvotes: 1
Views: 1086
Reputation: 1251
Not currently. But we are working on enabling this scenario to allow users to pick their choice of VMs. Unfortunately, I don't have a date yet but will update thread as we are closer to it.
Upvotes: 2