Updated PettingZoo+RLlib tutorial #19069
Conversation
Updated the tutorial and added link to the blog post by the PettingZoo team.
|
@richardliaw @sven1977 could you please merge this when you get a chance? |
|
@sven1977 @richardliaw could this please be merged? it's been 6 days |
rllib/examples/pettingzoo_env.py
Outdated
|
|
||
| # Train once | ||
| trainer.train() | ||
| env_name = "pistonball_v4" |
There was a problem hiding this comment.
Could we move most of the comments back here? They were quite useful for understanding, why we define certain config keys.
rllib/examples/pettingzoo_env.py
Outdated
| test_env = ParallelPettingZooEnv(env_creator({})) | ||
| obs_space = test_env.observation_space | ||
| act_space = test_env.action_space |
There was a problem hiding this comment.
I don't think we need this anymore. Just use:
multiagent:
policies: {"policy0"}
policy_mapping_fn: ...
rllib/examples/pettingzoo_env.py
Outdated
| }, | ||
| "gamma": 0.99, | ||
| } | ||
| return (None, obs_space, act_space, config) |
There was a problem hiding this comment.
same as above. I don't think it's needed anymore. Just move the "model" key (and its values) into the main config below. Same for gamma.
|
@sven1977 Per discussion with @jkterry1, we would prefer to move the pettingzoo example so that it links to the example in PettingZoo/tutorials. This way, we can maintain it to keep it compatible with both PettingZoo and RLlib. I've removed the outdated pettingzoo_env.py example and updated the docs accordingly for your review. There's also a Medium article on PettingZoo+RLlib that might be a useful link. |
Updated the tutorial and added link to the blog post by the PettingZoo team.
Why are these changes needed?
Checks
scripts/format.shto lint the changes in this PR.