I recently discovered the PyTorch Tensor and am very excited about its GPU support. However, it seems to support no interpolation algorithms as MATLAB / Numpy does. My application requires a pre-processing step using linear interpolation of the input data. It is OK now that this pre-processing is done in CPU using scipy.interpolate, but it will be nicer if there is something inside PyTorch Tensor that supports doing that operation inside GPU since they will be load into GPU eventually.

Since I am pretty new to PyTorch, I may be wrong that interpolation is an existing support operator (I did check the doc and found nothing about it). If not, do you guys think it is a good operation to be done in GPU? Is it something that you guys would like to include in PyTorch. If so, I can open a PR and start to working on it.

cc @ezyang @gchanan @zou3519 @mruberry @rgommers @heitorschueroff