I am trying to consolidate torch::where and torch::greater into a single function. I have bypass the binary output from torch::where to store the indexes of into a tensor object; however, the performance is not efficient. I have included my solution:
std::vector <int> idxs;
for(int x = 0; x < tensor.numel(); x++){
if (tensor[x].item<float>() > threshold)
idxs.push_back(x);
}