EdanToledo/Stoix
A research-oriented single-agent reinforcement learning framework implemented entirely in JAX for high-performance distributed training.
Not currently ranked — collecting fresh signals.
star history
Stoix provides optimized implementations of popular single-agent RL algorithms in JAX, enabling easy parallelization across devices via pmap and full compilation with jit for fast training. It targets researchers who want to quickly iterate on RL ideas, offering baseline implementations that can be tuned for specific environments. The codebase supports both JAX-native environments and external environments through its Sebulba system.
Frequently asked
- What is EdanToledo/Stoix?
- A research-oriented single-agent reinforcement learning framework implemented entirely in JAX for high-performance distributed training.
- Is Stoix open source?
- Yes — EdanToledo/Stoix is open source, released under the Apache-2.0 license.
- What language is Stoix written in?
- EdanToledo/Stoix is primarily written in Python.
- How popular is Stoix?
- EdanToledo/Stoix has 413 stars on GitHub.
- Where can I find Stoix?
- EdanToledo/Stoix is on GitHub at https://github.com/EdanToledo/Stoix.