Closed marty1885 closed 4 years ago
For example the following code does not work on OpenCL but works perfectly on CPU
reference_.view({(intmax_t)class_id}).assign(reference_.view({(intmax_t)class_id}) + sdr);
I've managed to track the bug down. The problem happens in the assign function and only happens when assigning from a 1D to a view of a 2D Tensor
assign
For example the following code does not work on OpenCL but works perfectly on CPU