pipeline.nodes is an array of objects, each object having the following properties:

  • name: name of the node, used in edge definitions edges:
  • type: type of the node, accepts “execution” or “deployment”
  • step: name of the step to be executed, defined in the same file
  • override: (optional) override values defined in the original step

Other than that, pipeline execution nodes behave like you would expect:

  • it will use the defaults from the original step
  • you can customize the parameters and inputs before starting the pipeline
  • you can override values when defining the pipeline using the override: syntax

Note that separate nodes in a pipeline can implement the same step multiple times.

# define "gather-dataset" and "train-model" steps above...
- pipeline:
    name: gatherer-pipeline
      - name: gather-node
        type: execution
        step: gather-dataset
      - name: train-node
        type: execution
        step: train-model
          image: tensorflow/tensorflow:1.13.1-py3
      - name: deploy-node
        type: deployment
        deployment: predict-digit
          - predict-digit
      - [gather-node.output.images*, train-node.input.dataset-images]
      - [gather-node.output.labels*, train-node.input.dataset-labels]
      - [train-node.output.model*, deploy-node.file.predict-digit.model]