A method for processing video data, including: deriving one or more motion vectors for a first set of sub-blocks belonging to a first dividing pattern of a current video block of a video; and performing, based on the one or more motion vectors, a conversion between the current video block and a bitstream of the video.