Authentication¶

Depending on the Configuration of A2ML, authentication will be with the A2ML Cloud or directly to providers.

A2ML Cloud¶

Authenticating with A2ML Cloud will soon allow you to have one account which manages all provider credentials. All providers can then be run without individual configuration.

Download Credentials¶

A2ML credentials can be downloaded as an auger.json file from your organization's https://app.auger.ai/<project_name>/settings.

$ export AUGER_CREDENTIALS=~/auger.json

Note

This would be a JSON string of your credentials.

The auger.json file can also be referred to by:

An environment variable called AUGER_CREDENTIALS_PATH

An attribute path_to_credentials inside config.yaml

Individual Providers¶

A2ML Cloud can manage the usage of Google Cloud AutoML or Azure AutoML for you. If set up on your own you must configure the underlying AutoML service corrrectly to be accessed from the server you are running.

To connect directly to any or all providers configure each individually.

Azure Provider¶

CLI Login¶

The Azure AutoML service allows browser login. Run any a2ml command and a login URL will open in the default browser.

To explicitly login to azure run.

$ a2ml auth login -p azure

JSON File¶

To login programmatically without the browser use the service principal credentials.

Note

Follow the above directions to create an azure.json file with values specific to your account.

azure.json¶

{
  "subscription_id":"",
  "directory_tenant_id":"",
  "application_client_id":"",
  "client_secret":""
}

$ export AZURE_CREDENTIALS=~/azure.json

Note

This would be a JSON string of your azure credentials.

The azure.json file can also be referred to by:

An environment variable called AZURE_CREDENTIALS_PATH - should contain a full path to the file

An attribute path_to_credentials inside config.yaml

Google Cloud Provider¶

If you haven't run Google Cloud AutoML, set up a service account and save the credentials to a JSON file which you store in your project directory. Then set up the GOOGLE_APPLICATION CREDENTIALS environment variable to point to the saved file. For example:

$ export GOOGLE_APPLICATION_CREDENTIALS="/Users/user/a2ml/automl.json"

For ease of use you can set up a default project ID to use with your project with the PROJECT_ID environment variable. For example:

$ export PROJECT_ID="automl-test-237311"

Detailed instructions for setting up Google Cloud AutoML.