Additional Configuration

Some additional model configuration must be specified, and some fields are optional. This will be explained below and it is relevant for both Python and R models.

Specifying prediction length constraints

Include min_prediction_length and max_prediction_length in your model configuration to define how many time periods your model can predict ahead. When users need predictions beyond your max_prediction_length, Chap automatically uses ExtendedPredictor to make iterative predictions (see supporting functionality).

Model Configuration Options

You can define configurable parameters in your MLproject file using user_options. This allows users to customize model behavior when running your model, without modifying the model code itself.

Schema structure

Each option in user_options has the following fields:

• title: Display name for the parameter
• type: One of string, integer, number, boolean, or array
• description: What the parameter does
• default: Optional default value. If omitted, the parameter is required

Example MLproject with user_options

name: my_model

docker_env:
  image: python:3.11

entry_points:
  train:
    parameters:
      train_data: str
      model: str
    command: "python train.py {train_data} {model}"
  predict:
    parameters:
      historic_data: str
      future_data: str
      model: str
      out_file: str
    command: "python predict.py {model} {historic_data} {future_data} {out_file}"

user_options:
  n_lag_periods:
    title: n_lag_periods
    type: integer
    default: 3
    description: "Number of lag periods to include in the model"
  learning_rate:
    title: learning_rate
    type: number
    description: "Learning rate for training (required)"

Providing configuration values

Configuration values can be provided via the --model-configuration-yaml CLI flag when running eval or other commands:

chap eval my_model data.csv results.nc --model-configuration-yaml config.yaml

The configuration YAML is a ModelConfiguration document with two top-level keys:

• user_option_values — a mapping whose keys must match the parameter names declared in your MLproject user_options
• additional_continuous_covariates — an optional list of extra continuous covariates (only meaningful when the template sets allow_free_additional_continuous_covariates: true)

user_option_values:
  n_lag_periods: 5
  learning_rate: 0.01
additional_continuous_covariates:
  - rainfall
  - mean_temperature

Validation rules

• The top-level YAML must contain only user_option_values and additional_continuous_covariates. Any other top-level key raises a validation error — flat YAMLs like n_lag_periods: 5 at the root are rejected.
• Options without a default value are required and must be provided under user_option_values
• Only options defined in user_options are allowed under user_option_values
• Values must match the specified type (e.g., integers for integer type)

Examples in the codebase

See the following examples that use user_options:

• naive_python_model_with_mlproject_file_and_docker
• web_based_model

Report Entry Point

In addition to the required train and predict entry points, a model may declare an optional report entry point that produces a PDF document describing the trained model (diagnostics, fitted-effect plots, posterior summaries, etc.). When defined, users can invoke it via the chap report CLI command.

Adding a report entry point

Add a third entry to entry_points in MLproject. The entry point receives the same historic_data and model files as predict, plus an out_file path the model must write the PDF to:

entry_points:
  train:
    parameters:
      train_data: str
      model: str
    command: "python train.py {train_data} {model}"
  predict:
    parameters:
      historic_data: str
      future_data: str
      model: str
      out_file: str
    command: "python predict.py {model} {historic_data} {future_data} {out_file}"
  report:
    parameters:
      historic_data: str
      model: str
      out_file: str
    command: "python report.py {model} {historic_data} {out_file}"

If the model has user options, declare model_config: str in the report parameters too — it is forwarded the same way as for train/predict.

Contract

• The model parameter is the trained-model artifact written by train.
• The historic_data parameter is a CSV in the standard Chap format.
• The script must write a PDF to out_file. Chap does not validate the contents.
• polygons.geojson is not passed to report.

Models without a report entry point are still valid — chap report will return an error message naming the model.