Building Container-based components

Container-based components provide the flexibility to integrate code written in any language into your pipeline, so long as you can execute that code in a Docker container.

If you are new to TFX pipelines, learn more about the core concepts of TFX pipelines.

Creating a Container-based Component

Container-based components are backed by containerized command-line programs. If you already have a container image, you can use TFX to create a component from it by using the create_container_component function to declare inputs and outputs. Function parameters:

• name: The name of the component.
• inputs: A dictionary that maps input names to types. outputs: A dictionary that maps output names to types parameters: A dictionary that maps parameter names to types.
• image: Container image name, and optionally image tag.
• command: Container entrypoint command line. Not executed within a shell. The command line can use placeholder objects that are replaced at compilation time with the input, output, or parameter. The placeholder objects can be imported from tfx.dsl.component.experimental.placeholders. Note that Jinja templates are not supported.

Return value: a Component class inheriting from base_component.BaseComponent which can be instantiated and used inside the pipeline.

Placeholders

For a component that has inputs or outputs, the command often needs to have placeholders that are replaced with actual data at runtime. Several placeholders are provided for this purpose:

• InputValuePlaceholder: A placeholder for the value of the input artifact. At runtime, this placeholder is replaced with the string representation of the artifact's value.

• InputUriPlaceholder: A placeholder for the URI of the input artifact argument. At runtime, this placeholder is replaced with the URI of the input artifact's data.

• OutputUriPlaceholder: A placeholder for the URI of the output artifact argument. At runtime, this placeholder is replaced with the URI where the component should store the output artifact's data.

Learn more about TFX component command-line placeholders.

Example Container-based Component

The following is an example of a non-python component that downloads, transforms, and uploads the data:

import tfx.v1 as tfx

grep_component = tfx.dsl.components.create_container_component(
    name='FilterWithGrep',
    inputs={
        'text': tfx.standard_artifacts.ExternalArtifact,
    },
    outputs={
        'filtered_text': tfx.standard_artifacts.ExternalArtifact,
    },
    parameters={
        'pattern': str,
    },
    # The component code uses gsutil to upload the data to Google Cloud Storage, so the
    # container image needs to have gsutil installed and configured.
    image='google/cloud-sdk:278.0.0',
    command=[
        'sh', '-exc',
        '''
          pattern="$1"
          text_uri="$3"/data  # Adding suffix, because currently the URI are "directories". This will be fixed soon.
          text_path=$(mktemp)
          filtered_text_uri="$5"/data  # Adding suffix, because currently the URI are "directories". This will be fixed soon.
          filtered_text_path=$(mktemp)

          # Getting data into the container
          gsutil cp "$text_uri" "$text_path"

          # Running the main code
          grep "$pattern" "$text_path" >"$filtered_text_path"

          # Getting data out of the container
          gsutil cp "$filtered_text_path" "$filtered_text_uri"
        ''',
        '--pattern', tfx.dsl.placeholders.InputValuePlaceholder('pattern'),
        '--text', tfx.dsl.placeholders.InputUriPlaceholder('text'),
        '--filtered-text', tfx.dsl.placeholders.OutputUriPlaceholder('filtered_text'),
    ],
)