figure of document error?

[laneeeee]

i found a mistake when i learn PagedAttention from document[https://docs.vllm.ai/en/latest/], i confused that the figure of "value" charater. there are HEAD_SIZE=128, BLCOK_SIZE=16, V_VEC_SIZE=8, WARP_SIZE=32 is known, why does the figure show that thd0 process v_vec16? according to my results, thd0 should process v_vec32