在Azure ML上开始训练模型

Question

I am totally new to Azure ML, and I need some help.

What I want:

• Build models in python
• Train model on Azure - (because I don't want to train models locally because of speed,...)
• Ideally, train more models in parallel so to speed up the time to develop the model.

I watch some videos, and read about it, but what I found is:

• Visual interface on Azure : It is not helpful because it only could read csv data -> no Images, Videos,...
• package in python to connect to azure (azure.core) for deploy and see metrics, but you still need to train model locally,...

Pleas, provide steps to do this

Answer 1

There is a good collection of sample notebooks how to use Azure ML on Github by Microsoft: https://github.com/Azure/MachineLearningNotebooks

For instance:

• Train on Azure ML managed compute
• Distributed deep learning with Tensorflow
• Training, hyperparameter tune, and deploy with Keras

And many more. Just look through the repo what fits best to your scenario.

Answer 2

Another potential resource, that focuses specifically on training for images (MNIST) is here:

https://docs.microsoft.com/en-us/azure/machine-learning/service/how-to-train-tensorflow

And if you scroll down about 70% of the page, there's even things on distributed training.