I am interested in writing a tutorial for using langchain with Shimmy, an API conversion tool allowing many popular reinforcement learning environments to be used natively with PettingZoo and Gymnasium.
Since there are already PettingZoo and Gymnasium tutorials/wrappers (https://python.langchain.com/en/latest/use_cases/agent_simulations/petting_zoo.html), it would not require significant new code, but I think it would be a really helpful feature to show people that they can for example load DM Control/DM Lab environments in the exact same way. OpenSpiel provides text-rendered board games, which would be a great feature to show off with a chatbot.
Making this issue just to confirm it is something that the developers would be interested, but I am planning to work on the tutorial later today and if it's not possible to add directly to this repo I will add it to Shimmy (I am the project manager of Shimmy and PettingZoo).
I would be interested in fleshing out integration on PettingZoo's side as well, if there are any extra features which we could add in order to better support langchain we would love to do so. If there is interest in adding training library support I am also interested in working on tutorials to have langchain load a trained model using standard libraries like stable-baselines3, RLlib, or a lighter library like CleanRL.
vowelparrot
commented
1 year ago
I am interested in writing a tutorial for using langchain with Shimmy, an API conversion tool allowing many popular reinforcement learning environments to be used natively with PettingZoo and Gymnasium.
Since there are already PettingZoo and Gymnasium tutorials/wrappers (https://python.langchain.com/en/latest/use_cases/agent_simulations/petting_zoo.html), it would not require significant new code, but I think it would be a really helpful feature to show people that they can for example load DM Control/DM Lab environments in the exact same way. OpenSpiel provides text-rendered board games, which would be a great feature to show off with a chatbot.
Making this issue just to confirm it is something that the developers would be interested, but I am planning to work on the tutorial later today and if it's not possible to add directly to this repo I will add it to Shimmy (I am the project manager of Shimmy and PettingZoo).
I would be interested in fleshing out integration on PettingZoo's side as well, if there are any extra features which we could add in order to better support langchain we would love to do so. If there is interest in adding training library support I am also interested in working on tutorials to have langchain load a trained model using standard libraries like stable-baselines3, RLlib, or a lighter library like CleanRL.