def _count_message_tokens()

in yourbench/utils/inference/inference_tracking.py [0:0]

• 13 lines of code
• 6 McCabe index (conditional complexity)

def _count_message_tokens(messages: List[Dict[str, str]], encoding: tiktoken.Encoding) -> int:
    """Counts tokens in a list of messages, approximating OpenAI's format."""
    num_tokens = 0
    # Approximation based on OpenAI's cookbook: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynb
    # This might not be perfectly accurate for all models/providers but is a reasonable estimate.
    tokens_per_message = 3
    tokens_per_name = 1

    for message in messages:
        num_tokens += tokens_per_message
        for key, value in message.items():
            if value:
                num_tokens += _count_tokens(str(value), encoding)
            if key == "name":
                num_tokens += tokens_per_name
    num_tokens += 3
    return num_tokens