Cuda_time vs. cuda_time_total in torch.autograd.profiler.profile

1

When doing:

a = torch.randn(10,10).cuda()
b = torch.randn(10,100).cuda()
with torch.autograd.profiler.profile(use_cuda=True) as prof:
    ret = a.mm(b)

perf = prof.total_average()

What does each one mean?