Okay... Okay... They took too much to accept the question, that I found a solution.
The solution was make the global position be the current one... So the lerp function works
self.transform.origin.z = lerp(self.transform.origin.z, player_posi_z, smooth_cam * delta)
self.transform.origin.y = lerp(self.transform.origin.y, player_posi_y + 5, smooth_cam * delta)