Super suit migration example #1091
Conversation
| This function takes an environment, a module, and a parameter, and creates a new environment with an observation | ||
| space and observations modified based on the provided module and parameter. | ||
|
|
||
| Parameters: |
There was a problem hiding this comment.
If you could have the docstrings be more similar to the syntax in conversions.py and other wrappers in PettingZoo that would be good. Also I don't know the wrapper super well but I think my_module and my_param may not be the correct names. In general using GPT to write this stuff is pretty risky because it could very well just be completely made up and I don't have the time to look through all of the specifics to ensure it's correct. If you can do so yourself and double check then that's great but I'm hesitant to include too much details because it could be incorrect. Look elsewhere throughout the repo to see how the formatting is done for the docstrings.
And for the example format, see chess.py as we have an example using the >>> format which gets tested under the doctests with pytest.
There was a problem hiding this comment.
Ok thanks for the advice, I'll fix the docstring format and write them from scratch.
There was a problem hiding this comment.
Agree, definitely not use the output of CGPT directly for docstrings.
| class BaseWrapper(PettingzooWrap): | ||
| def __init__(self, env: AECEnv | ParallelEnv): | ||
| """ | ||
| Creates a wrapper around `env`. Extend this class to create changes to the space. |
There was a problem hiding this comment.
Also this looks incorrect saying class BaseWrapper(PettingZooWrap), why would it use OrderEnforcingWrapper as the base class? And I'd honestly have to think more about what the base class should be, it's tough cause a lot of the functions like step() and reset() have different return types and aren't uniform between parallel and AEC.
There was a problem hiding this comment.
I haven't really looked into the wrappers for the moment, this is just as it was on SuperSuit I can try to find a better base class for BaseWrapper
|
Also run the pre-commit hooks as that will give you some feedback about docstrings adhering to the right format. |
|
Looks fairly sensible so far, seems like these are just files taken from SS with docstrings added. For type hinting, I would recommend something like the way Gymnasium does it, but ultimately this is Elliot's call. I'll wait for a more fleshed out version of the PR before doing a more in depth review. Ultimately, as long as the tests are comprehensive and they pass I don't see anything wrong. |
Agreed, and we’ve added some better type hinting modeled after that for other internal files in pettingzoo, so that should be easy to replicate looking at env.py and stuff |
|
Jets pr got merged which was an example of doubt SuperSuit style wrappers, you should try to follow it as an example |
Perfect , thanks a lot, I'll try to find some time and start with the migration |
|
@GiovanniGrotto For your reference, it's #1105 :) |
Description
This is a try of migrating clip_reward_v0 and color_reduction_v0 from SuperSuit basic_wrappers.py to PettingZoo.
It is a work in progress so there is a lot of things that needs to be fixed but I would like to know if at least I'm in the good track.
In particular:
I'm having some trouble with type hinting, in fact some of them could be wrong, since sometimes it is not very clear what a function should accepts, is there a way or a best practice to make this easier or at least cleaner?
I'm open to any type of comment and suggestions about how to make this migration cleaner.
Type of change
Screenshots
Checklist:
pre-commitchecks withpre-commit run --all-files(seeCONTRIBUTING.mdinstructions to set it up)pytest -vand no errors are present.pytest -vhas generated that are related to my code to the best of my knowledge.