I'm using Nvidia Nsight system for profiling ML applications in PyTorch.
and I would like to see just the range that I want to focus on.
(I mean not notation like NVTX, but turn on and off the profiler itself.)
I use both GUI and CLI.
For CLI,
below command works well, with --capture-range=cudaProfilerApi this option.
torch.cuda.cudart().cudaProfilerStop()
torch.cuda.cudart().cudaProfilerStart()
But in GUI, it shows me the whole range of the application.
Does anyone know how to use cudaProfilerApi in Nsight GUI? (As far as I searched, it has only manual turn-off and wall clock time-based control, not code base)