Function evaluate_during_training

utils/trainingEvaluator.py:346–380

Convenience function to call from training loop. Args: model: The model being trained tokenizer: The tokenizer test_data: Test dataset or path to test data device: Device to run evaluation on save_dir: Directory to save results epoch: Cur

(model, tokenizer, test_data, device="cpu", 
                           save_dir="./evaluation_results", epoch=None)

# Convenience function to use in training loop
def evaluate_during_training(model, tokenizer, test_data, device="cpu", 
                           save_dir="./evaluation_results", epoch=None):
    """
    Convenience function to call from training loop.
    
    Args:
        model: The model being trained
        tokenizer: The tokenizer
        test_data: Test dataset or path to test data
        device: Device to run evaluation on
        save_dir: Directory to save results
        epoch: Current epoch number (optional, for naming)
    
    Returns:
        dict: Evaluation metrics
    """
    # Create evaluator
    if epoch is not None:
        save_dir = f"{save_dir}/epoch_{epoch}"
    
    evaluator = ModelEvaluator(save_dir=save_dir)
    
    # Run evaluation
    metrics = evaluator.evaluate_model(
        model=model,
        tokenizer=tokenizer,
        test_data=test_data,
        device=device,
        batch_size=8,
        save_results=True,
        plot_confusion_matrix=True,
        analyze_errors=True
    )
    
    return metrics


# Example usage in training script