About the Q-shift operation in the code

[kkkkkk123-ops]

in the jit_func in Class VRWKV_SpatialMix_V6 why we need to -x after the shift_func? It seems not -x in the _inner_forward in Class VRWKV_ChannelMix when calculate xx.
#Class VRWKV_SpatialMix_V6
def jit_func(self, x, patch_resolution):
# Mix x with the previous timestep to produce xk, xv, xr
B, T, C = x.size()

    xx = self.shift_func(x, self.shift_pixel, patch_resolution=patch_resolution, 
                         with_cls_token=self.with_cls_token) - x
    xxx = x + xx * self.time_maa_x  # [B, T, C]

#Class VRWKV_ChannelMix
def _inner_forward(x):
xx = self.shift_func(x, self.shift_pixel, patch_resolution=patch_resolution,
with_cls_token=self.with_cls_token)