gym wrapper, train_muzero_with_gym_env

[rosnu]

Hello,
1.
Could someone verify if entry_type == "train_muzero_with_gym_env" (setting at the end of the script) is working?
I tried it with:
python3 -u zoo/classic_control/cartpole/config/cartpole_efficientzero_config.py

Error:
/content/LightZero/lzero/worker/muzero_collector.py in collect(self, n_episode, train_iter, policy_kwargs, collect_with_pure_policy)
329 policy_kwargs = {}
330 temperature = policy_kwargs['temperature']
--> 331 epsilon = policy_kwargs['epsilon']
332
333 collected_episode = 0
KeyError: 'epsilon'

entry_type == "train_muzero" seems to be working fine.
I'm more of an "enthusiast" than a professional, so I would prefer to continue with a working example.

Is it beneficial to build a custom environment from scratch according to the LiteZero schema compared to using a Gym/Gymnasium environment and a wrapper? I can't find any configuration example for any train_xxxx_with_gym_env in the zoo.

Thank you!

[puyuan1996]

Hello, regarding the first issue you mentioned, we have already fixed it in this commit. As for the second issue, you can refer to our documentation: Customizing Environments Tutorial. If you encounter any problems while customizing the environment, feel free to start a discussion here. Thank you again for your attention and support!

[r-wedeen]

Hi, I am also using train_muzero_with_gym_env to wrap my gymnasium environment with LightZeroEnvWrapper. What should my create_config look like? I was not able to find this information in the Customizing Environments Tutorial mentioned above nor in the Customizing Configuration Files Tutorial. Thanks!

[puyuan1996]

Hello, if your environment has been correctly configured according to the documentation, the create_config should be similar to the configuration in this link. Of course, you may need to add extra parameters for the custom environment in env filed of the main_config. If you have any questions, feel free to ask at any time. Thank you for your support!