https://github.com/microsoft/DeepSpeed-Kernels/blob/ec12a38c18f4bb7d31d3bbabb39786eb2ba6b063/dskernels/inf_flash_attn/blocked_flash/flash_fwd_launch_template.h#L140 and https://github.com/microsoft/DeepSpeed/blob/ce5e56a82eef66888456e75c45b5ed1214cfc54e/deepspeed/inference/v2/kernels/ragged_ops/blocked_flash/blocked_flash.py#L56 is inconsistent.
https://github.com/microsoft/DeepSpeed-Kernels/blob/ec12a38c18f4bb7d31d3bbabb39786eb2ba6b063/dskernels/inf_flash_attn/blocked_flash/flash_fwd_launch_template.h#L140
and https://github.com/microsoft/DeepSpeed/blob/ce5e56a82eef66888456e75c45b5ed1214cfc54e/deepspeed/inference/v2/kernels/ragged_ops/blocked_flash/blocked_flash.py#L56
is inconsistent.