Setting up Google Cloud

This document will guide you through the process of configuring your machine to submit jobs to Google’s Cloud AI Platform with caliban cloud.


Caliban will eventually support other Cloud providers like AWS, but Google Cloud will remain a great starting option. Google Cloud gives you $300 of credit, so you can get started immediately with caliban cloud.

By the end you’ll be able to complete the final step of the “Getting Started with Caliban” tutorial and train an ML model on AI Platform.

These are the steps we’ll cover:

To make it through, you’ll need to know:

Create a Cloud Account

To submit jobs to AI Platform you need to create a Google Cloud account with an active “project”. Every job you submit to Cloud will be associated with this project.

Visit the Google Cloud Console and click “Select a Project”:

Select a Project

Click “New Project” in the dialogue that appears:

New Project

Give your project a memorable name like “totoro-lives” and click Create. This should take you to your new project’s dashboard.

Note the Project ID in the “Project info” panel:

Project ID

Caliban will use this project ID to submit jobs to the correct project in Cloud.

Add the following line to the file ~/.bashrc or ~/.bash_profile on your machine, to make the ID available to Caliban:

export PROJECT_ID=<your-project-id>


If you don’t know what this means, see this page for a tutorial on environment variables. We’ll remind you at the end of the tutorial which variables you’ll need.

Activate Free Cloud Trial

Every new Google Cloud project comes with a free $300 credit. To activate this, click “Activate” at the top right of your new Cloud account’s console and follow the prompts.

Activate Billing

You’ll have to set up a billing account as well.

The system will ask you for a credit card to verify your identity, but if you use up the entire credit it won’t automatically charge you. You can decide at that point whether you’d like to continue or not.

Enable AI Platform and Container Registry

Google Cloud has a dizzying number of products. To submit jobs with caliban cloud, you’ll need to activate just these two:

Follow the instructions at this link to enable the Container Registry API by selecting the project you created above and clicking “Continue”.

Click this link to Enable the AI Platform Jobs API by clicking “Enable API” and waiting for the spinner to stop.

Install the Cloud SDK

The final step is to install the Google Cloud SDK on your machine.

Visit the Google Cloud SDK installation page for a full set of installation instructions. Here is the distilled version:

If you didn’t do it during installation, you’ll need to initialize the SDK with this command:

gcloud init

When you see this output:

You are logged in as: [].

This account has a lot of projects! Listing them all can take a while.
 [1] Enter a project ID
 [2] Create a new project
 [3] List projects
Please enter your numeric choice:

Enter 1, then type in your project ID you noted earlier. (You should have saved it as an environment variable called $PROJECT_ID).

If you’d like to set a default zone, anything beginning with us-central1 is a great choice. us-central1 has the most capability of any region.


the Cloud SDK is quite powerful, and gives you access to Cloud buckets and all sorts of Google services. You might want to peruse the full set of SDK documentation once you’ve got everything working.

Configure Docker Authentication

To submit a job with caliban cloud, Caliban needs to push Docker images to the Container Registry service that you enabled earlier. To allow Docker to push to the Container Registry, run this command at a terminal:

gcloud auth configure-docker

You should see output that includes the text Adding credentials for all GCR repositories..

Test your Environment

To check if your SDK installation was successful, run gcloud auth list in your terminal. You should see your email address listed as the active account:

[totoro@totoro ~]$ gcloud auth list
    Credentialed Accounts

To set the active account, run:
    $ gcloud config set account `ACCOUNT`

As a final step, confirm that you’ve set the following environment variables. (If you set a custom region above, add it here as a $REGION variable).

export REGION="us-central1"
export PROJECT_ID="research-3141"

If you have all of this, you’re set!

Train a model in Cloud

Now that you have a working Cloud configuration and a new project, you can use caliban cloud to submit jobs to Cloud AI platform.

The “Getting Started with Caliban” tutorial ends with a nice demo that has you training models in Cloud. Head over to the tutorial and complete the final step to train a digit-classifying neural network on AI Platform.