tensorflow_makefile icon indicating copy to clipboard operation
tensorflow_makefile copied to clipboard

Published 20 hours ago • verified publisher icon petewarden

uint64_t support in eigen_spatial_convolutions.h

Open mayank0403 opened this issue 5 years ago • 0 comments

Hi, I wanted to use SpatialConvolution() function with Tensors of uint64_t type.

If I try invoking SpatialConvolution() with Tensors of int/float/double types, then the function works fine, but as soon as I try any of uint64_t/int64_t, I get a compilation error saying that DataMapper does not have member nonStandardPatches() and SubMapper does not have member loadCoeffStandard().

This error doesn't seem to come with tensors having int scalers.

Does SpatialConvolution() not support 64-bit integers at the moment?

mayank0403 avatar Apr 11 '19 20:04 mayank0403