void NN_sub(Tensor *out, const Tensor *a, const Tensor *b)
void NN_sub_4d(Tensor *out, const Tensor *a, const Tensor *b)
void NN_sub_2d(Tensor *out, const Tensor *a, const Tensor *b)
void NN_sub_1d(Tensor *out, const Tensor *a, const Tensor *b)
void NN_sub_3d(Tensor *out, const Tensor *a, const Tensor *b)