Function process_theorem

evaluate.py:245–346

Process a theorem file or directory for evaluation. Args: models: Dictionary of models for different evaluation types. file_path (str): Path to the file or directory to evaluate. eval_type (str): Type of evaluation to perform. retry_limit (int): Number of re

(models, file_path: str, eval_type: str, retry_limit: int,
                    target_fps: int = None, use_parent_folder_as_topic: bool = False,
                    output_folder: str = None)

def process_theorem(models, file_path: str, eval_type: str, retry_limit: int,
                    target_fps: int = None, use_parent_folder_as_topic: bool = False,
                    output_folder: str = None) -> tuple[str, dict]:
    """
    Process a theorem file or directory for evaluation.

    Args:
        models: Dictionary of models for different evaluation types.
        file_path (str): Path to the file or directory to evaluate.
        eval_type (str): Type of evaluation to perform.
        retry_limit (int): Number of retry attempts.
        target_fps (int, optional): Target frames per second for video processing.
        use_parent_folder_as_topic (bool, optional): Use parent folder name as topic.
        output_folder (str, optional): Directory to store output files.

    Returns:
        tuple[str, dict]: Tuple of file name and evaluation results.
    """
    ext_map = {
        'text': ('.txt', '.srt'),
        'video': ('.mp4', '.mkv')
    }

    # Handle single file evaluation
    if os.path.isfile(file_path):
        file_ext = os.path.splitext(file_path)[1].lower()
        file_name = os.path.basename(file_path)

        if eval_type == "text" and file_ext in ext_map['text']:
            return file_name, evaluate_text_file(models['text'], file_path, retry_limit)
        elif eval_type == "video" and file_ext in ext_map['video']:
            if use_parent_folder_as_topic:
                topic_name = os.path.basename(os.path.dirname(file_path))
            else:
                topic_name = None
            topic_name = process_topic_name(topic_name)
            return file_name, evaluate_video_file(models['video'], file_path, None, topic_name, target_fps, output_folder)
        elif eval_type == "image" and file_ext in ext_map['video']:
            if use_parent_folder_as_topic:
                topic_name = os.path.basename(os.path.dirname(file_path))
            else:
                topic_name = None
            topic_name = process_topic_name(topic_name)
            return file_name, evaluate_sampled_images(models['image'], file_path, topic_name, num_chunks=10, output_folder=output_folder)
        elif eval_type == "all":
            raise ValueError("Evaluation type 'all' is not supported for a single file. Try passing a folder with both a video and a subtitle file.")
        else:
            raise ValueError(f"File type of {file_path} does not match evaluation type {eval_type!r}")

    # Handle directory evaluation
    theorem_dir = file_path
    all_files = os.listdir(theorem_dir)

    # Look for transcript files, prioritizing .srt over .txt if both exist
    transcript_file_candidates = [f for f in all_files if f.endswith(ext_map['text']) and not f.endswith('_scene_outline.txt')]
    srt_files = [f for f in transcript_file_candidates if f.endswith('.srt')]
    txt_files = [f for f in transcript_file_candidates if f.endswith('.txt')]