I believe you can slightly change the flash attention kernel to implement the same kernel of this page attention, since both of them work on the key/value cache

I believe you can slightly change the flash attention kernel to implement the same kernel of this page attention, since both of them work on the key/value cache at block level.