vllm_omni.metrics.utils ¶

count_tokens_from_outputs(engine_outputs: list[Any]) -> int