Method exist_prefill

fastdeploy/worker/gpu_model_runner.py:290–296

check whether prefill stage exist

(self)

                logger.exception("Exception in async output loop: %s", e)

    def exist_prefill(self):
        """
        check whether prefill stage exist
        """
        if envs.ENABLE_V1_KVCACHE_SCHEDULER:
            return self.exist_prefill_flag
        return (self.share_inputs["seq_lens_encoder"] > 0).any().cpu().numpy().item()

    @property
    def is_sleeping(self):