ai_library

ai_library is a Python package that exposes model training, inference, metrics recording, cron scheduling, and package-based configuration management.

Package overview

• model selection and all other parameters are done via the config file.

  • train writes to the config file model (path), infer tries to load a model of that type.

• ai_library.validate_config(config_path=None)

  • Loads and validates the package default ai_library/config.yaml if no path is provided.
  • Returns the parsed configuration dictionary.

• ai_library.update_config(config_path=None, updates=...)

  • Updates the package config file by default.
  • Supports a dictionary, a [key, value] pair, or a list of [key, value] pairs.

• ai_library.show_config(config_path=None)

  • Prints the current configuration to stdout.

• ai_library.train()

  • Loads configuration from save_file path of the package config.
  • Reads data, builds the selected pipeline, and trains the model.

• ai_library.infer()

  • Loads package configuration from save_file path from config.
  • Loads a saved model and performs inference.

• ai_library.record

  • Use ai_library.record.main() or run the module directly to start metric collection. Run in background.

• ai_library.add_to_cron() / ai_library.remove_from_cron()

  • Manage cron scheduling for recurring training runs.
  • Adds or removes a cron job that runs python3 -m ai_library.codebase.setup.train.