Context
I think we all recognize brax environments are useful for very fast RL training. How do we create custom envs which are useful?
Setup
- Prerequisites
- Repo or package links
- Environment details
Core steps
- Define environment state and action space.
- Implement reset and step transitions.
- Register the environment.
- Connect it to your training loop.
Common issues
- Symptom: …
- Cause: …
- Fix: …
References
- Link 1
- Link 2
Next actions
- Follow-up task 1
- Follow-up task 2