What is the implementation in the lines?

[TicTacToePP]

xz_s = xz.chunk(self.nframes, dim=-1)
xz_s = torch.stack(xz_s,dim=-1)
xz_s = xz_s.flatten(-2)

I tested these lines and found xz_s is the same as original xz for the starting and ending elements, which confuses me. Could you please help explain the lines? thanks!