Feature request: bilinear upsampling #28
Description
Hi I already saw the upsample_bilinear_op in the list of wrapped kernels; could you make it accessible at a higher level? This is relevant for speeding up UNets and similar, which is in turn important for semantic segmentation tasks.