completed part two, had to use cpu and get_tensor_affine. need a clean

Source diff could not be displayed: it is too large. Options to address this: view the blob.