Apps · PROJECT NOTES
AgentGym-RL
A multi-turn reinforcement-learning stack that connects AgentGym environment servers to rollout generation and policy optimization built on verl.
What can you do with it?
Train and evaluate agents over multiple environment turns, including experiments that progressively lengthen the interaction horizon.
What should you know before using it?
This adds training infrastructure to AgentGym rather than introducing an entirely separate set of environments. The documented recipe pins CUDA 12.4, PyTorch 2.4 and Python 3.10, with a Linux x86-64 attention wheel. Compute and environment servers are separate prerequisites. Code is MIT.
What is verified here?
This entry was reviewed against original project material. We have not independently run the software or reproduced its results. AI-development claims are author-reported; check upstream for current versions, setup requirements, and licenses.
- Resource type
- Environment framework
- Provenance
- Development provenance not established
- Reviewed
- Platforms
- Consult upstream documentation