UniRL
Train multimodal AI models with one unified reinforcement learning loop
All Skills in This Repository (4)
Pure Emerald Level Indicators
Tencent-Hunyuanpr-workflow
Create and repair UniRL pull request bodies and titles for GitHub review workflows.
Tencent-Hunyuanadd-model-bundle
Scaffold UniRL model bundle support for diffusion or autoregressive packages.
Tencent-Hunyuancode-standards
Enforce coding standards and review-driven workflows for Python, PyTorch, shell scripts, and configs.
Tencent-Hunyuanpr-workflow
Create and repair GitHub pull requests following UniRL template and CI rules.
Frequently Asked Questions
FAQPage SchemaHow to install UniRL?â–¼
Run `npx skills add Tencent-Hunyuan/UniRL --all -g -y` in your terminal to install all skills in this suite globally.
What is UniRL used for?â–¼
UniRL is a reinforcement learning framework that post-trains multimodal models like Stable Diffusion, WAN video, and Qwen using one shared training loop.
Which models does UniRL support?â–¼
It supports image diffusion models like SD3 and FLUX, video models like WAN and HunyuanVideo, and language models like Qwen3 and Qwen-VL.
Does UniRL work with AI coding agents?â–¼
Yes. It ships CLAUDE.md guidelines and SKILL.md files that teach agents how to add model bundles and manage pull requests correctly.
Can I run UniRL training recipes without writing code?â–¼
Yes. You launch pre-built YAML recipes with a single command, and the framework handles rollout, scoring, and distributed training automatically.
Related Repositories in Software Engineering
View All in Software Engineering→
openclawopenclaw
Run a personal AI assistant across your devices and chat apps
obrasuperpowers
Gives coding agents a disciplined workflow from idea to merged code
facebookreact
AI agent skills for building, testing, and porting React core