Tensorcore identification while running inference

Hello,
Currently, I am inferencing through tensorflow using onnx .
Is there any way to find out how many tensorcores have been used while inferencing?
I have checked in nsight system but not able to find any information.
Please help.

Hi,

Tensor Core supports IMMA and HMMA operations.
Please find Nsight Compute to get the utilization of these operations:

tensor_precision_fu_utilization (HMMA)
tensor_precision_int_utilization (IMMA)

Thanks.

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.