Tencent-Hunyuan/UniRL

[RFC] Community refresh of main-repository documentation 3 days ago

documentation help wanted

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

[RFC] Extend UniRL's on-policy distillation support (Welcome Contributing!) 7 days ago

help wanted

Tencent-Hunyuan/UniRL

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

reward(imagebind): the default mode scores no prompt, and it is the mode that collapses video under RL 9 days ago

help wanted

Tencent-Hunyuan/UniRL

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

[RFC] Validate composite reward capabilities before eager model loading 10 days ago

enhancement help wanted

Tencent-Hunyuan/UniRL

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

refactor(sglang): make Sample sampling parameters authoritative 12 days ago

help wanted

Tencent-Hunyuan/UniRL

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

reward: no way to score a component without feeding it into the gradient (monitor-only rewards) 16 days ago

help wanted

Tencent-Hunyuan/UniRL

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

feat(meta-init): resolve Hugging Face repo IDs before sharded loading 26 days ago

help wanted

Tencent-Hunyuan/UniRL

918

UniRL is a Framework for Unified Multimodal Model Reinforcement Learning

Python

#ai-infrastructure#reinforcement-learning#sglang#vllm

Open Issues Need Help