huggingface / huggingface/deep-rl-class
[HANDS-ON BUG] Colab + Python 3.12 issues in Unit 1 (Box2D, gymnasium[box2d], and repo publish step)
- Dominant language
- MDX
- Stars
- 5k
- Forks
- 811
- PR merge metrics
- No merged PRs in 30d
Description
# Describe the bug
When completing Unit 1 of the Deep RL course on Google Colab (Python 3.12), multiple cells fail due to dependency and platform incompatibilities, particularly around Box2D environments and the model publishing step.
Specifically:
* gymnasium[box2d] fails to install on Colab with Python 3.12
* Box2D environments such as LunarLander, BipedalWalker, and CarRacing raise `ModuleNotFoundError: No module named 'Box2D'`
* box2d-py and pygame==2.1.3 fail to build from source
* Several later cells implicitly assume Box2D availability, causing repeated failures even after earlier workarounds
* The “Publish our trained model on the Hub” step fails with `403 Forbidden` unless the user manually authenticates, but this is not clearly documented and Colab does not always expose “Tools → Secrets”
I was able to complete the unit by replacing the Box2D environments with CartPole-v1, which works correctly and still satisfies the learning objectives.
Please share your notebook link so that we can reproduce the error:
https://colab.research.google.com/github/huggingface/deep-rl-class/blob/master/notebooks/unit1/unit1.ipynb
# Material
* Did you use Google Colab?
Yes
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.