Experiment Configuration Manager

Build a configuration management system that creates independent experiment variants from a base configuration template without affecting the original configuration.

Requirements

Your system should support the following operations:

Base Configuration Setup

Create a base configuration for a machine learning experiment with the following structure:

• MODEL section containing:
  • TYPE: string specifying the model architecture
  • LEARNING_RATE: float value for training
  • BATCH_SIZE: integer value
• DATA section containing:
  • TRAIN_PATH: string path to training data
  • VAL_PATH: string path to validation data
• TRAINING section containing:
  • EPOCHS: integer number of training epochs
  • USE_GPU: boolean flag

Creating Experiment Variants

Implement functionality to create independent experiment configurations from the base configuration. Each variant should be completely independent - modifications to one variant should not affect the base configuration or other variants.

Create three experiment variants:

1. high_lr_experiment: Increases learning rate by 10x from base
2. large_batch_experiment: Doubles the batch size from base
3. extended_training_experiment: Triples the number of epochs from base

Verification

Your implementation must verify independence by:

• Modifying each variant as specified above
• Confirming the base configuration remains unchanged after creating and modifying all variants
• Confirming each variant has only its specific modification

Test Cases

• Creating a variant and modifying it leaves the base configuration unchanged @test
• Multiple variants can be created from the same base without interfering with each other @test
• Modifications to nested configuration values in a variant do not affect the base @test

Implementation

@generates

API

def create_base_config():
    """
    Creates and returns a base configuration with MODEL, DATA, and TRAINING sections.
    Returns a configuration object suitable for creating independent variants.
    """
    pass

def create_experiment_variant(base_config, variant_name, modifications):
    """
    Creates an independent copy of the base configuration and applies modifications.

    Args:
        base_config: The base configuration object to copy from
        variant_name: String name for this experiment variant
        modifications: Dict of configuration keys and new values to apply

    Returns:
        An independent configuration object with modifications applied
    """
    pass

def verify_independence(base_config, variants):
    """
    Verifies that all variants are independent from the base configuration.

    Args:
        base_config: The original base configuration
        variants: List of variant configurations to check

    Returns:
        Boolean indicating whether all variants are truly independent
    """
    pass

Dependencies { .dependencies }

yacs { .dependency }

Provides configuration management support.