Index

Types and functions

InferenceReport.InferenceType

The object of a report created by this package:

  • algorithm: Algorithm that created the samples (currently assumes it is a PT struct). Optional, nothing if unspecfied.
  • chains: Samples, stored in a Chains struct.
source
InferenceReport.PostprocessContextType

Information passed to the processors:

  • inference: The Inference on which the report will be based on.
  • output_directory: The resolved location of the output report.
  • generated_markdown: Vector of Strings used internally to create the markdown page.
  • generated_dict: Generated dictionary used internally to create a JSON object for indexing purposes.
source
InferenceReport.ReportOptionsType

Options to control report:

  • max_moving_plot_iters: Maximum number of iterations in moving_pair_plot. Set to zero to avoid creating such movie.

    Creating movies are a bit slow, so in interactive mode the default is zero, and 100 in headless mode.

  • target_name: Name of the target. If nothing, will attempt to auto-detect a name via multiple dispatch of the functions target_name or pigeons_target_name.
  • render: If true, the report webpage's md files are rendered into html files.
  • view: If true, the report webpage will be opened automatically.
  • exec_folder: The directory where the report will be created. A unique folder in results/all will be created by default and symlinked in results/latest.
  • reproducibility_command: A command to instruct users how to reproduce.

    At the moment implicitly assumes Pigeons will be used, and running these commands should return the Inputs.

source
InferenceReport.pigeons_target_descriptionMethod
pigeons_target_description(target)

By default, dispatch to pigeons_target_description(target, Val(Symbol(pigeons_target_name(target)))), to handle "vague types" such as TuringLogPotential and StanLogPotential.

source