we have trained a humanoid transformer with large-scale reinforcement learning in simulation and deployed it to the real world zero-shot