Import your inference data from S3 to Aporia

Article Content

    Want to monitor your ML models without touching your serving code? We just released a cool new tool for you!

    Aporia Importer is an open-source utility to import your inference data from cloud storage providers such as Amazon S3 and others into Aporia.

    The tool currently supports CSV and Parquet files.

    Installation

    pip install aporia-importer[all]

    Usage

    Copy the following YAML config file and adjust it to your needs:

    # Data source configuration
    
    # Specify the URL/path to your data here (either a local file or s3)
    # Glob patterns are supported, but all files must be of the same format
    source: s3://my-bucket/data.parquet
    
    # Specify the format of the files in the source pattern above
    # Supported formats: csv, parquet
    format: parquet
    
    # SDK configuration (these values are passed directly to the Aporia SDK)
    token: <YOUR APORIA TOKEN>
    environment: production
    
    # Model configuration
    
    # You can find the model_id after creating a model in the Aporia dashboard
    model_id: my-model-a2f4
    
    # Define your model version schema here
    # See: https://auth.prod.aporia.com/login?state=hKFo2SA1NnV3VS04b09iRW9BY1FuMXczXzdPQVlBTnhfMDNZc6FupWxvZ2luo3RpZNkgYzFkQ21SYWQwODJRcWk0N0lQeWt3NzdCOXJMd1BiLXWjY2lk2SBNb0VhN1cwdTM3OG0zWnFkazQ4bnViWjdWY1pZODNHMA&client=MoEa7W0u378m3Zqdk48nubZ7VcZY83G0&protocol=oauth2&error_message=&invite_token=&mode=&redirect_uri=https%3A%2F%2Fapp.aporia.com%2Fv1%2Fauth%2Fcallback&response_type=code&scope=openid%20profile%20email/docs/getting-started/concepts/#model-version-schema)
    model_version:
      name: v1
      type: binary
      predictions:
        will_buy_insurance: boolean
      features:
        Age: numeric
        Driving_License: boolean
        Region_Code: categorical
        Annual_Premium: numeric
        Vintage: numeric
        Vehicle_Damage: boolean

    Now you can run:

    aporia-importer config.yaml

    As mentioned above, this utility is completely open-source, so if need to join datasets, read different file formats, or pretty much anything else – you can easily customize it for your needs.

    If you have feature requests or any other suggestions, please let us know!

    Start Monitoring Your Models in Minutes