#2929th of 53 in Training and fine-tuningFor science, health and research, robotics
rlinf
Distributed RL training framework for robot VLA policies and LLM agents
- Stars
- 5.6k
- License
- Apache-2.0
- Last commit
- Nov 2026
- Last release
- Oct 2026
- Language
- Python
3,113 on 21 Apr 20265,632 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
RLinf is a Python RL training infrastructure for embodied and agentic AI. It runs PPO, GRPO, SFT, DAgger and related methods on VLA policies (OpenPI π₀/π₀.₅, GR00T, OpenVLA-OFT), world models and LLMs, in simulators such as LIBERO, RoboTwin and Isaac Lab, and on real robots. It supports several accelerator backends and Megatron-based actor training, and installs from PyPI.
Who it is for: Robotics and RL researchers post-training VLA models or LLM agents on GPUs
Strengths
- Covers simulated and real-robot RL, including teleoperation, SFT and deployment pipeline
- Wide model support: π₀/π₀.₅, GR00T, OpenVLA-OFT, Cosmos3, Wan2.2, Moonlight-16B
- Eight accelerator backends, including AMD ROCm, Ascend CANN, Moore Threads and MetaX
- Adopted by Isaac Lab v3.0.0; installable from PyPI; docs in English and Chinese
Weaknesses
- Each model and simulator integration has its own setup; the feature set is large
- No Docker or compose files detected in the repository
- Aimed at robotics and RL researchers, not general LLM fine-tuning
- Real-robot workflows depend on specific supported hardware such as Franka
What it needs
- GPU required
- Docker
- Needs PyTorch, Megatron, SGLang, Simulators (LIBERO, RoboTwin, Isaac Lab)
- Models: OpenPI π₀ / π₀.₅, π₀-FAST, OpenVLA-OFT, GR00T-N1.6/N1.7, Cosmos3
Quick answers
Is rlinf open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Can you run rlinf with Docker?
Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.
Does rlinf need a GPU?
Yes, it needs a GPU.
Also in Training and fine-tuning
See all 53| Rank | Project | Score |
|---|---|---|
| 1 | unslothDesktop and web app to run and fine-tune local models | 88 out of 100 |
| 2 | peftPython library for LoRA and other parameter-efficient fine-tuning of Transformers models | 88 out of 100 |
| 3 | deepspeedPyTorch library for distributed training of very large models with ZeRO | 86 out of 100 |
| #4 | verlReinforcement learning training library for LLM post-training with PPO and GRPO | 86 out of 100 |
| #5 | trlPost-training library with SFT, DPO, GRPO and KTO trainers | 86 out of 100 |