CUs assignment status

1

Hi @jtuyls /@mak ,

Is there any way we can get the CUs assignment status? like model1 has assigned to CU1 and model2 has assigned to CU2

Thanks and Regards, Raju

2

@kvaju.454 You can use dexplorer -s to check the DPU cores’ utilization however I am not aware of a method to check which model was assigned to which CU. Note that this is for the DNNDK runtime (which we are currently using on Pynq). We will be moving to the VART flow shortly and there you might be able to use the DEBUG_DPU_RUNNER environment variable to check which model/VART Runner is assigned to which CU. I expect it to be incremental though, first runner uses CU 0, next one CU 1.

3

Hi @jtuyls ,

Thanks for the information and will wait for VART workflow release.

Thanks and Regards, Raju