Autopilot Training

Make sure you saw this link for preprocessing first

Create AutoML job using the console or CLI or Python

Using Python

Click here

Configure data for AutoML job

  • Set the location of the data set,
  • Select the target attribute that I want the model to predict: in this case, it’s the ‘y’ column showing if a customer accepted the offer or not,
  • Set the location of training artifacts.
input_data_config = [{
      'DataSource': {
        'S3DataSource': {
          'S3DataType': 'S3Prefix',
          'S3Uri': 's3://{}/{}/input'.format(bucket,prefix)
        }
      },
      'TargetAttributeName': 'y'
    }
  ]

output_data_config = {
    'S3OutputPath': 's3://{}/{}/output'.format(bucket,prefix)
  }

Create AutoML job

auto_ml_job_name = 'automl-dm-' + timestamp_suffix

import boto3
sm = boto3.client('sagemaker')
sm.create_auto_ml_job(AutoMLJobName=auto_ml_job_name,
                      InputDataConfig=input_data_config,
                      OutputDataConfig=output_data_config,
                      RoleArn=role)

Using CLI

Click here

Set data config and create AutoML job

aws sagemaker create-auto-ml-job \
--auto-ml-job-name my-automl-job \
--input-data-config '[
        {
            "DataSource": {
                "S3DataSource": {
                    "S3DataType": "S3Prefix",
                    "S3Uri": "s3://<bucket>/<prefix>/input"
                }
            },
            "CompressionType": "None",
            "TargetAttributeName": "y"
        }
    ]'
--output-data-config '{
        "KmsKeyId": "",
        "S3OutputPath": "s3://<bucket>/<prefix>/output"
    }'
--role-arn "arn:aws:iam::<account-id>:role/<role-name-with-path>"

Using Console

Click here

Updated on 28 Feb 2020


Related content: