Pierre TASSEL

When you use `SubprocVecEnv` as a wrapper to have multiple actors in parallel, the environments automatically call the `reset()` method after the end of an episode. I think the `compute_returns()`...

Hi, The behavior you observed is normal. As the environment contains illegal action depending on the state, you have to sample for the legal action vector. Please refer to this...

Thanks a lot for the interest and the compliment ;) > I didn't saw the discussion, thanks for the hint - it looks like it can help me. I've included...

Hello, Thank you, it's nice to hear it is useful :) And thank you for reviewing the code, that's important. > You are first making sure that the length of...

The observation provided by the environment contains both a boolean array indicating if the action is legal or not and the "real" observation ```python self.observation_space = gym.spaces.Dict({ "action_mask": gym.spaces.Box(0, 1,...

Hi @karthik7194, First of all, sorry for the late answer :( This project is a simple training approach using PPO to train an agent on the Job-Shop Scheduling Environment we...

You are right! Thanks for the notification. I will fix this ASAP

Hi @smita-09 Thank you for your kind words :) It seems there is some issue with the registration of the environment. Could you try to add those lines at the...

Hi, Unfortunately it's not possible right now But you can easily implement it by changing the check for the end of the job https://github.com/prosysscience/JSSEnv/blob/f24556433152316420ce7dbe5dac18fa17cb49b7/JSSEnv/envs/jss_env.py#L387

Hi, This seems to be an error related to Ray (RLLib), check this issue for help: https://github.com/ray-project/ray/issues/11395