Config API

Config is PETsARD’s internal configuration management class, responsible for parsing configuration dictionaries and building module execution sequences.

ℹ️

Internal Use Only: Config is primarily used internally by Executor. Users should interact with PETsARD through Executor instead of using Config directly.

Class Architecture

Basic Usage

from petsard.config import Config

# Create Config from dictionary
config_dict = {
    "Loader": {
        "load_data": {
            "filepath": "data.csv"
        }
    },
    "Synthesizer": {
        "generate": {
            "method": "sdv",
            "model": "GaussianCopula"
        }
    }
}

config = Config(config_dict)

# Config is typically used internally by Executor
from petsard import Executor
exec = Executor('config.yaml')  # Executor creates Config internally

Constructor

Syntax

Config(config: dict)

Parameters

Parameter	Type	Required	Description
`config`	`dict`	Yes	Configuration dictionary containing all module configurations

Return Value

Returns a Config instance with parsed configuration.

Configuration Structure

Config expects a dictionary with the following structure:

{
    "ModuleName": {
        "experiment_name": {
            "parameter1": "value1",
            "parameter2": "value2"
        }
    }
}

Example

config_dict = {
    "Loader": {
        "load_data": {"filepath": "data.csv"}
    },
    "Synthesizer": {
        "method_a": {"method": "sdv", "model": "GaussianCopula"},
        "method_b": {"method": "sdv", "model": "CTGAN"}
    }
}

Core Attributes

Attribute	Type	Description
`config`	`queue.Queue`	Queue of instantiated Adapters
`module_flow`	`queue.Queue`	Queue of module names
`expt_flow`	`queue.Queue`	Queue of experiment names
`sequence`	`list`	Module execution order
`yaml`	`dict`	Processed configuration dictionary

Automatic Processing

1. Module Ordering

Config automatically arranges modules in the correct order:

Loader → Preprocessor → Splitter → Synthesizer →
Postprocessor → Constrainer → Evaluator → Reporter

2. Splitter Expansion

When Splitter configuration includes num_samples > 1, Config automatically expands it:

# Input
"Splitter": {
    "split_data": {
        "train_split_ratio": 0.8,
        "num_samples": 3
    }
}

# Automatically expanded to
"Splitter": {
    "split_data_[3-1]": {"train_split_ratio": 0.8, "num_samples": 1},
    "split_data_[3-2]": {"train_split_ratio": 0.8, "num_samples": 1},
    "split_data_[3-3]": {"train_split_ratio": 0.8, "num_samples": 1}
}

3. Cartesian Product Generation

Multiple experiments generate all combinations:

config_dict = {
    "Loader": {
        "load_v1": {...},
        "load_v2": {...}
    },
    "Synthesizer": {
        "method_a": {...},
        "method_b": {...}
    }
}

# Generates 4 execution paths:
# load_v1 → method_a
# load_v1 → method_b
# load_v2 → method_a
# load_v2 → method_b

Configuration Validation

Config validates experiment names to ensure they don’t use reserved patterns:

# ✓ Valid experiment names
"experiment_name"
"load_data"
"method_v2"

# ✗ Invalid experiment names (reserved pattern)
"experiment_[1]"    # Raises ConfigError
"load_data_[test]"  # Raises ConfigError

Integration with Executor

Config is typically used through Executor:

from petsard import Executor

# Executor internally:
# 1. Loads YAML file
# 2. Creates Config instance
# 3. Uses Config to build execution queue
exec = Executor('config.yaml')
exec.run()

Notes

Internal Use: Config is primarily for internal use by Executor
Recommended Practice: Use YAML configuration files with Executor rather than Config directly
Naming Restrictions: Experiment names cannot use _[xxx] pattern (reserved for system use)
Automatic Processing: Splitter multi-sample configurations are automatically expanded
Execution Order: Module order is automatically determined by Config
Memory Usage: Large experiment combinations may consume significant memory