verl-rl-training - verl: LLM RL Training Framework
Flexible RL training library for large language models
Tags
Updated: 2026-06-30Capabilities
Typical Inputs
Typical Outputs
What this skill does
- train models
- run rollout
- compute reward
- configure algorithms
- swap backends
- execute multi-turn rollout
- train vision-language models
- enable LoRA
Inputs
- Base model
- Training dataset
- Reward function
- Training config
Outputs
- Trained model checkpoints
- Training metrics
Requirements
- Python 3.8+
- verl>=0.3.0
- torch>=2.0.0
- ray>=2.41.0
- vllm>=0.8.2
- transformers>=4.40.0
- GPU cluster
- HuggingFace model