Inquiries on monitoring GPU utilization through dcgm

Developer Tools Other Tools System Management and Monitoring (NVML)
1

Hello, I’m using the writing # Allowing multiple processes to watch DCGM profile fields

We are monitoring the item ‘DCGM_FI_PROF_GR_ENGINE_ACTIVE’ and there is no change in the value of the tensorflow mnist code using GPUs, and there is no change in the value of the code (GitHub - manycoresoft/gpu-stress-test) for testing.

I was wondering if someone can help with the following questions:

  1. Do I have to calculate the collected values separately?

  2. Do you limit the use of multi-instance GPUs when enabled?