I was conducting a few experiments to see if changing the order of matrix multiplications in linear attention would yield the same results.