Android App TensorFlow Google Cloud ML

Question

I am studying how to use TensorFlow together with Google Cloud ML on an Android App. I already found this post. As far as I understand from this post and what I already found on google I always have to deploy a trained model (i.e. the Tensorflow Graph?) first? There is no way to use Google Cloud ML to send in data values and train a TensorFlow Model in this Cloud?

So the only way is: deploy a model, send data values to the google cloud and receive a prediction?

Thanks

Answer 1

There is an API for running training jobs in the Cloud. For an example, see the training quickstart.

Note that there is not currently a straightforward way to perform streaming training directly using the CloudML service, but you could get creative with PubSub, Cloud DataFlow, or other technologies.

Were you looking for a streaming solution or is the job-based solution sufficient for your needs?