[Bug] AttributeError: 'TimeLimitWrapper' has no attribute 'agent' when converting control mode in replay_trajectory for new gymnasium

[luokuang2001]

I am trying to convert a dataset collected via Motion Planning (using pd_joint_pos) into pd_ee_delta_pos control mode for training a Diffusion Policy.

When running the replay_trajectory script to perform this conversion, the script crashes with an AttributeError.

I run the following command to replay and convert the control mode:

python -m mani_skill.trajectory.replay_trajectory \
  --traj-path ./trajectory.h5 \
  -c pd_ee_delta_pos \
  -o rgb \
  --save-traj

The error log is:

[zatpds]

Hi, got the same error. I solved it by patching it locally.
put
ori_env = ori_env.unwrapped
env = env.unwrapped
in the conversion.py file

[luokuang2001]

Hi, got the same error. I solved it by patching it locally. put ori_env = ori_env.unwrapped env = env.unwrapped in the conversion.py file

Thanks a lot !
After executing ori_env = ori_env.unwrapped and env = env.unwrapped in conversion.py, the resulting JSON file is missing specific episode information.
Crucially, the step() function must be called on the wrapped environment.

[StoneT2000]

I think this might be one of the gymnasium bugs that inevitably will crop due to our upgrade to gym >1.0 in #940

@luokuang2001 @zatpds can you share what the output of this is

pip list | grep gymnasium

[luokuang2001]

I think this might be one of the gymnasium bugs that inevitably will crop due to our upgrade to gym >1.0 in #940

@luokuang2001 @zatpds can you share what the output of this is

pip list | grep gymnasium

Yes, it turns out to be an issue with the Gym version.

[milong26]

I met the problem when following replay demos

python -m mani_skill.trajectory.replay_trajectory \
  --traj-path ~/.maniskill/demos/PickCube-v1/motionplanning/trajectory.h5 \
  --use-first-env-state -c pd_ee_delta_pos -o state \
  --save-traj --num-envs 1 -b physx_cpu --count 1 --record-video

Then I got

Replaying traj_0: : 0step [00:02, ?step/s]Traceback (most recent call last):
  File "/home/user/miniconda3/envs/mskill/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/home/user/miniconda3/envs/mskill/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/home/user/working_folder/ManiSkill/mani_skill/trajectory/replay_trajectory.py", line 627, in <module>
    main(parse_args())
  File "/home/user/working_folder/ManiSkill/mani_skill/trajectory/replay_trajectory.py", line 606, in main
    _, replay_result = _main(
  File "/home/user/working_folder/ManiSkill/mani_skill/trajectory/replay_trajectory.py", line 464, in _main
    replay_result = replay_cpu_sim(
  File "/home/user/working_folder/ManiSkill/mani_skill/trajectory/replay_trajectory.py", line 280, in replay_cpu_sim
    ori_env.set_state_dict(ori_env_states[0])
AttributeError: 'TimeLimitWrapper' object has no attribute 'set_state_dict'
Replaying traj_0: : 0step [00:02, ?step/s]
  0%|                                                                               | 0/1 [00:02<?, ?step/s]

I tested pip list | grep gymnasium, find gymnasium 1.2.3

"gymnasium>=0.29.1" in setup.py so I pip uninstall gymnasium and reinsatll by pip install gymnasium==0.29.1

It works :)
It seems gymnasium==0.29.1 still works