Submitting tuning jobs in Google's AI Platform

Now we are ready to use Google's AI Platform to perform hyperparameter training. You may download everything you need from the GitHub repository for this chapter. For the AI Platform code in this section, you can refer to the gcptuningwork file in this chapter's folder in the GitHub repository for the book.

In the cloud, we have access to powerful machines that can speed up our search process. Overall, the approach we will leverage is very similar to what we saw in the previous section about submitting a local Python script training job. We will use the tf.compat.v1.flag method to handle user input or flags. The rest of the script follows a similar structure, with the exception of data handling, because we will use TFRecord instead of ImageGenerator and a conditional flag for the distributed training strategy.

Since the tuning job is submitted to AI Platform from a remote node (that is, your local compute environment...