Change tensor.data to tensor.detach() due to
pytorch/pytorch#6990 (comment)
tensor.detach() is more robust than tensor.data.

I figured out a way to get the (x,y,z) data points for each frame from one hand previously. but im not sure how to do that for the new holistic model that they released. I am trying to get the all landmark data points for both hands as well as parts of the chest and face. does anyone know how to extract the holistic landmark data/print it to a text file? or at least give me some directions as to h

AutoNonVariableTypeMode, which is used in our C++ backward kernels as in https://github.com/pytorch/vision/blob/9e6bdbabcf2a39563d69dad81d8fddef5800ae3f/torchvision/csrc/ops/autograd/roi_align_kernel.cpp#L88, is being deprecated and will be removed in the 1.10 release of PyTorch, see https://github.com/pytorch/pytorch/blob/master/docs/cpp/source/notes/inference_mode.rst#migration-guide-from-auto

It's normal to have many map() calls, and some of them can take a few minutes,
it would be nice to have a description on the progress bar.

Alternative solution:
Print the description before/after the map() call.