Closed andrewzhang505 closed 1 year ago
Base: 79.86% // Head: 79.83% // Decreases project coverage by -0.02%
:warning:
Coverage data is based on head (
4866dba
) compared to base (0b1d3fc
). Patch coverage: 0.00% of modified lines in pull request are covered.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
Overall this is very good, I like it!
Here's a few changes:
1) Put this into examples/notebooks
2) Notebook should not reference anything from sf_examples folder like train_gym_env, it should be self-contained because this will show how this can work with any environment, not just Gym
.
Other than that, I think we can keep it. As for the logging, I think you can manipulate it with logging module settings in Python, but users will probably want logging anyway? I think the way you did it now is good!
It looks like there is an issue with using multiprocessing with a function defined in IPython (https://stackoverflow.com/questions/41385708/multiprocessing-example-giving-attributeerror). This is giving me an error when I try to define the make_gym_env_func
inside the notebook instead of importing it from sf_examples
Hmm if we can't use Multiprocessing in this notebook why bother using it at all... This kind of defies the whole purpose. I guess we can still have some examples that work in serial mode, like envpool.
Okay, how about we merge this as is, but maybe you can put a short comment in the notebook explaining the situation with custom functions. Also move it to examples/notebooks.
BTW here's an interesting link about this: https://stackoverflow.com/a/65001152/1645784
Thank you, I think it looks good!
Added example notebook for training lunar lander and uploading to the hub. I didn't tune the hyperparameters so it doesn't train that well, but it does work. Where should I put the notebook?