CLI Reference

Complete reference for all Pilz command-line commands.

Commands Overview

flowchart TD
    A[pilz] --> B[train]
    A --> C[eval]
    A --> D[infer]
    A --> E[create-dc]

    B --> B1[Train models]
    C --> C1[Evaluate + metrics]
    D --> D1[Predictions only]
    E --> E1[Generate DataCard]

    style A fill:#e0f0ff
    style B fill:#ccffcc
    style C fill:#ffff99
    style D fill:#ffff99
    style E fill:#ccffcc```

![Diagram](images/cli_reference_1.svg)

## train

Train Pilz models.

```bash
pilz train --datacard <file.yaml> --trainsettings <file.yaml>

Options

Option	Description
--datacard	Path to DataCard YAML file
--trainsettings	Path to TrainSettings YAML file

Example

pilz train --datacard iris_dc.yaml --trainsettings train.yaml

eval

Evaluate trained models and generate ROC curves.

pilz eval --datacard <file.yaml> --evalsettings <file.yaml>

Options

Option	Description
--datacard	Path to DataCard YAML file
--evalsettings	Path to EvalSettings YAML file

Example

pilz eval --datacard iris_dc.yaml --evalsettings eval.yaml

Output Files

• {target}_roc.html - ROC curve per class
• all_roc.html - Combined ROC curves
• multi_class_result.html - Per-class accuracy
• {out_file} - Prediction scores (CSV/Parquet)

infer

Run inference only (no metrics).

pilz infer --datacard <file.yaml> --evalsettings <file.yaml>

Options

Option	Description
--datacard	Path to DataCard YAML file
--evalsettings	Path to EvalSettings YAML file

Example

pilz infer --datacard iris_dc.yaml --evalsettings eval.yaml

Use Cases

• Fast prediction on new data
• Batch scoring in production
• When you only need scores, not metrics

create-dc

Generate a DataCard from a dataset.

pilz create-dc --src <path> --out <file.yaml>

Options

Option	Description
--src	Path to CSV or Parquet file
--out	Output DataCard YAML file (optional, will prompt)

Example

pilz create-dc --src data.csv --out datacard.yaml

Interactive Prompts

1. Select target column
2. For each feature, choose type (categorial/numerical)
3. Confirm missing value handling

Quick Reference

Command	Purpose
pilz train --datacard X --trainsettings Y	Train models
pilz eval --datacard X --evalsettings Y	Evaluate with metrics
pilz infer --datacard X --evalsettings Y	Fast predictions
pilz create-dc --src X --out Y	Generate DataCard