name: TFX - Analyze model
description: |
  Runs Tensorflow Model Analysis. https://www.tensorflow.org/tfx/model_analysis/get_started
  TensorFlow Model Analysis allows you to perform model evaluations in the TFX pipeline, and view resultant metrics and plots in a Jupyter notebook. Specifically, it can provide:
  * metrics computed on entire training and holdout dataset, as well as next-day evaluations
  * tracking metrics over time
  * model quality performance on different feature slices
inputs:
- {name: Model,           type: GCSPath, description: GCS path to the model which will be evaluated.} # type: {GCSPath: {path_type: Directory, data_type: Exported TensorFlow models dir}}
- {name: Evaluation data, type: GCSPath,              description: GCS path of eval files.} # type: {GCSPath: {data_type: CSV}}
- {name: Schema,          type: GCSPath, description: GCS json schema file path.} # type: {GCSPath: {data_type: TFDV schema JSON}}
- {name: Run mode,        type: String,     default: local,               description: whether to run the job locally or in Cloud Dataflow.}
- {name: GCP project,     type: GCPProjectID, default: '',                  description: 'The GCP project to run the dataflow job, if running in the `cloud` mode.'}
- {name: Slice columns,   type: String,                                   description: Comma-separated list of columns on which to slice for analysis.}
- {name: Analysis results dir, type: GCSPath,   description: GCS or local directory where the analysis results should be written.} # type: {GCSPath: {path_type: Directory}}
outputs:
- {name: Analysis results dir, type: GCSPath,   description: GCS or local directory where the analysis results should were written.} # type: {GCSPath: {path_type: Directory}}
- {name: MLPipeline UI metadata, type: UI metadata}
implementation:
  container:
    image: gcr.io/ml-pipeline/ml-pipeline-dataflow-tfma:57d9f7f1cfd458e945d297957621716062d89a49
    command: [python2, /ml/model_analysis.py]
    args: [
      --model,    {inputValue: Model},
      --eval,     {inputValue: Evaluation data},
      --schema,   {inputValue: Schema},
      --mode,     {inputValue: Run mode},
      --project,  {inputValue: GCP project},
      --slice-columns, {inputValue: Slice columns},
      --output,   {inputValue: Analysis results dir},
    ]
    fileOutputs:
      Analysis results dir: /output.txt
      MLPipeline UI metadata: /mlpipeline-ui-metadata.json