Traceback (most recent call last):
  File "/Users/mattstaines/robot-bench/trainer/rollout.py", line 136, in <module>
    main()
  File "/Users/mattstaines/robot-bench/trainer/rollout.py", line 52, in main
    replay = record(env, policy, obs, seconds=args.seconds, seed=args.seed,
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattstaines/robot-bench/trainer/rollout.py", line 87, in record
    a, _, _ = policy.act(obs[None], rng, deterministic=deterministic)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/mattstaines/robot-bench/trainer/ppo.py", line 21, in act
    x = self.norm(obs)
        ^^^^^^^^^^^^^^
  File "/Users/mattstaines/robot-bench/trainer/nets.py", line 29, in __call__
    return np.clip((x - self.mean) / np.sqrt(self.var + 1e-8), -self.clip, self.clip)
                    ~~^~~~~~~~~~~
ValueError: operands could not be broadcast together with shapes (1,30) (22,)