Hi, I run a multiple process application on CUDA. And now I want to monitor per process’s utilization on GPU. I know that nivida-smi can do that by using “nvidia-smi pmon”, however, after I using the MPS, I find that except for the mps process, all rest process’ gpu utilization is zero. Could you please tell me which tool I can use to monitor per process’s performance when MPS is running, like nvidia-smi do? Thanks very much!
Have you found any solution? I’ve faced this problem too.