Configuration Manager

Build a simple configuration manager for a machine learning training system that supports flexible access patterns and nested configuration structures.

Requirements

Create a module that manages training configurations with the following capabilities:

Configuration Creation and Access

The system should support creating hierarchical configurations with nested structures. Configurations should be accessible using both attribute-style notation (e.g., config.TRAIN.BATCH_SIZE) and dictionary-style notation (e.g., config['TRAIN']['BATCH_SIZE']).

Type Safety

The configuration system should enforce type safety by only allowing specific data types: strings, integers, floats, booleans, tuples, lists, and None values. Any attempt to store unsupported types should be prevented.

Test Cases

• Given an empty configuration, when system settings are added with NUM_GPUS=4 and NUM_WORKERS=8, then these values should be retrievable using both attribute-style access (config.SYSTEM.NUM_GPUS) and dict-style access (config['SYSTEM']['NUM_WORKERS']) @test

• Given a configuration with nested structure (TRAIN.OPTIMIZER.LEARNING_RATE=0.001 and TRAIN.OPTIMIZER.MOMENTUM=0.9), when accessing via mixed access patterns (config.TRAIN['OPTIMIZER'].LEARNING_RATE), then all values should be correctly retrievable @test

• Given a configuration, when attempting to store a value of type 'set' or 'complex', then an error should be raised indicating invalid type @test

Implementation

@generates

API

class ConfigNode:
    """A configuration container that supports both attribute and dictionary access patterns."""

    def __init__(self, init_dict=None):
        """
        Initialize a configuration node.

        Args:
            init_dict: Optional dictionary to initialize the configuration
        """
        pass

    def __getattr__(self, name):
        """Support attribute-style access (e.g., config.KEY)."""
        pass

    def __setattr__(self, name, value):
        """Support attribute-style assignment with type validation."""
        pass

    def __getitem__(self, key):
        """Support dictionary-style access (e.g., config['KEY'])."""
        pass

    def __setitem__(self, key, value):
        """Support dictionary-style assignment with type validation."""
        pass

Dependencies { .dependencies }

yacs { .dependency }

Provides configuration management support for building the ConfigNode class.

@satisfied-by