How to get the queuing and inference time of each user in offline batch inference mode?

[fengxijia]

Given the local model, prompts, how to get the queuing and inference time of each user in offline batch inference mode? Particularly, streaming synchronous and streaming asynchronous generation.