verl: Volcano Engine Reinforcement Learning for LLMs

verl is a flexible, efficient and production-ready RL training library for large language models (LLMs).

verl is the open-source version of HybridFlow: A Flexible and Efficient RLHF Framework paper.

verl is flexible and easy to use with:

• Easy extension of diverse RL algorithms: The hybrid-controller programming model enables flexible representation and efficient execution of complex Post-Training dataflows. Build RL dataflows such as GRPO, PPO in a few lines of code.

• Seamless integration of existing LLM infra with modular APIs: Decouples computation and data dependencies, enabling seamless integration with existing LLM frameworks, such as FSDP, Megatron-LM, vLLM, SGLang, etc

• Flexible device mapping: Supports various placement of models onto different sets of GPUs for efficient resource utilization and scalability across different cluster sizes.

• Ready integration with popular HuggingFace models

verl is fast with:

• State-of-the-art throughput: SOTA LLM training and inference engine integrations and SOTA RL throughput.

• Efficient actor model resharding with 3D-HybridEngine: Eliminates memory redundancy and significantly reduces communication overhead during transitions between training and generation phases.

News

• [2025/05] verl will be presented at GOSIM x PyTorch Day 2025. See you in Paris!
• [2025/04] We will give a tutorial about latest post-training techniques and programming guide for verl at ICLR 2025. See you in Singapore!
• [2025/03] verl v0.3.0.post1 is released! See release note for details.
• [2025/03] DAPO is the open-sourced SOTA RL algorithm that achieves 50 points on AIME 2024 based on the Qwen2.5-32B pre-trained model, surpassing the previous SOTA achieved by DeepSeek's GRPO (DeepSeek-R1-Zero-Qwen-32B). DAPO's training is fully powered by verl and the reproduction code is publicly available now.
• [2025/03] We introduced the programming model of verl at the vLLM Beijing Meetup and verl intro and updates at the SGLang-LMSYS Org Meetup in Sunnyvale mid-March.

more...

• [2025/02] verl v0.2.0.post2 is released!
• [2025/01] [Doubao-1.5-pro](https://team.doubao.com/zh/special/doubao_1_5_pro) is released with SOTA-level performance on LLM & VLM. The RL scaling preview model is trained using verl, reaching OpenAI O1-level performance on math benchmarks (70.0 pass@1 on AIME).
• [2025/03] We will present verl(HybridFlow) at EuroSys 2025. See you in Rotterdam!
• [2025/02] We presented verl in the Bytedance/NVIDIA/Anyscale Ray Meetup. See you in San Jose!
• [2024/12] verl is presented at Ray Forward 2024. Slides available here
• [2024/10] verl is presented at Ray Summit. Youtube video available.
• [2024/12] The team presented Post-training LLMs: From Algorithms to Infrastructure at NeurIPS 2024. Slides and video available.
• [2024/08] HybridFlow (verl) is accepted to EuroSys 2025.

Key Features

• FSDP and Megatron-LM for training.
• vLLM, SGLang and HF Transformers for rollout generation.
• Compatible with Hugging Face Transformers and Modelscope Hub: Qwen-2.5, Llama3.1, Gemma2, DeepSeek-LLM, etc
• Supervised fine-tuning.
• Reinforcement learning with PPO, GRPO, ReMax, REINFORCE++, RLOO, PRIME, etc.
  • Support model-based reward and function-based reward (verifiable reward)
  • Support vision-language models (VLMs) and multi-modal RL
• Flash attention 2, sequence packing, sequence parallelism support via DeepSpeed Ulysses, LoRA, Liger-kernel.
• Scales up to 70B models and hundreds of GPUs.
• Experiment tracking with wandb, swanlab, mlflow and tensorboard.

Upcoming Features

• Roadmap https://github.com/volcengine/verl/issues/710
• DeepSeek 671b optimizations with Megatron v0.11 https://github.com/volcengine/verl/issues/708
• Multi-turn rollout optimizations
• Environment interactions

Getting Started

Documentation

Quickstart:

• Installation
• Quickstart
• Programming Guide

Running a PPO example step-by-step:

• Data and Reward Preparation
  • Prepare Data for Post-Training
  • Implement Reward Function for Dataset
• Understanding the PPO Example
  • PPO Example Architecture
  • Config Explanation
  • Run GSM8K Example

Reproducible algorithm baselines:

• PPO, GRPO, ReMax

For code explanation and advance usage (extension):

• PPO Trainer and Workers
  • PPO Ray Trainer
  • PyTorch FSDP Backend
  • Megatron-LM Backend
• Advance Usage and Extension
  • Ray API design tutorial
  • Extend to Other RL(HF) algorithms
  • Add Models with the FSDP Backend
  • Add Models with the Megatron-LM Backend
  • Deployment using Separate GPU Resources

Blogs from the community

• 使用verl进行GRPO分布式强化学习训练最佳实践
• HybridFlow veRL 原文浅析
• 最高提升20倍吞吐量！豆包大模型团队发布全新 RLHF 框架，现已开源！

Performance Tuning Guide

The performance is essential for on-policy RL algorithm. We have written a detailed performance tuning guide to help you optimize performance.

Upgrade to vLLM v0.8.2

verl now supports vLLM>=0.8.2 when using FSDP as the training backend. Please refer to this document for the installation guide and more information. Please avoid vllm 0.7.x, which contains bugs that may lead to OOMs and unexpected errors.

Use Latest SGLang

SGLang is fully supported with verl, and SGLang RL Group is working extensively on building unique features, including multi-turn agentic RL, VLM RLHF, server-based RL, and partial rollout. Please refer to this document for the installation guide and more information.

Citation and acknowledgement

If you find the project helpful, please cite:

• HybridFlow: A Flexible and Efficient RLHF Framework
• A Framework for Training Large Language Models for Code Generation via Proximal Policy Optimization

@article{sheng2024hybridflow,
  title   = {HybridFlow: A Flexible and Efficient RLHF Framework},
  author  = {Guangming Sheng and Chi Zhang and Zilingfeng Ye and Xibin Wu and Wang Zhang and Ru Zhang and Yanghua Peng and Haibin Lin and Chuan Wu},
  year    = {2024},
  journal = {arXiv preprint arXiv: 2409.19256}
}

verl is inspired by the design of Nemo-Aligner, Deepspeed-chat and OpenRLHF. The project is adopted and contributed by Bytedance, Anyscale, LMSys.org, Alibaba Qwen team, Shanghai AI Lab, Tsinghua University, UC Berkeley, UCLA, UIUC, University of Hong Kong, ke.com, All Hands AI, ModelBest, OpenPipe, JD AI Lab, Microsoft Research, StepFun, Amazon, Linkedin, Meituan, Camel-AI, OpenManus, Prime Intellect, NVIDIA research, Baichuan, and many more.

Awesome work using verl

• TinyZero: a reproduction of DeepSeek R1 Zero recipe for reasoning tasks
• DAPO: the fully open source SOTA RL algorithm that beats DeepSeek-R1-zero-32B
• SkyThought: RL training for Sky-T1-7B by NovaSky AI team.
• simpleRL-reason: SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild
• Easy-R1: Multi-modal RL training framework
• OpenManus-RL: LLM Agents RL tunning framework for multiple agent environments.
• deepscaler: iterative context scaling with GRPO
• rllm: async RL training with verl-pipeline
• PRIME: Process reinforcement through implicit rewards
• RAGEN: a general-purpose reasoning agent training framework
• Logic-RL: a reproduction of DeepSeek R1 Zero on 2K Tiny Logic Puzzle Dataset.
• Search-R1: RL with reasoning and searching (tool-call) interleaved LLMs
• ReSearch: Learning to Reason with Search for LLMs via Reinforcement Learning
• DeepRetrieval: Hacking Real Search Engines and retrievers with LLMs via RL for information retrieval
• Code-R1: Reproducing R1 for Code with Reliable Rewards
• Skywork-OR1: Skywork open reaonser series
• ToRL: Scaling tool-integrated RL
• cognitive-behaviors: Cognitive Behaviors that Enable Self-Improving Reasoners, or, Four Habits of Highly Effective STaRs
• PURE: Credit assignment is the key to successful reinforcement fine-tuning using process reward model
• MetaSpatial: Reinforcing 3D Spatial Reasoning in VLMs for the Metaverse
• DeepEnlighten: Reproduce R1 with social reasoning tasks and analyze key findings
• DeepResearcher: Scaling deep research via reinforcement learning in real-world environments
• self-rewarding-reasoning-LLM: self-rewarding and correction with generative reward models
• critic-rl: LLM critics for code generation
• VAGEN: Training VLM agents with multi-turn reinforcement learning
• AdaRFT: Efficient Reinforcement Finetuning via Adaptive Curriculum Learning
• Trust Region Preference Approximation: A simple and stable reinforcement learning algorithm for LLM reasoning.
• DQO: Enhancing multi-Step reasoning abilities of language models through direct Q-function optimization
• FIRE: Flaming-hot initiation with regular execution sampling for large language models
• Rec-R1: Bridging Generative Large Language Models and Recommendation Systems via Reinforcement Learning
• all-hands/openhands-lm-32b-v0.1: A strong, open coding agent model, trained with multi-turn fine-tuning

Contribution Guide

Contributions from the community are welcome! Please check out our project roadmap and good first issues to see where you can contribute.

Code Linting and Formatting

Warning

We are immigrating to ruff as the linter and formatter and pre-commit as the managing tool.

If your branch is based on a previous commit using yapf and pylint, simply merging might trigger overwhelming linting errors, while you are only expected to resolve ones in the files related to your PR.

To resolve this issue, please try the following workaround to only include the files you really changed in the PR:

1. In your branch, fix linting and format with ruff: ruff check --fix && ruff-format
2. Squash into a new single commit: git reset --soft $(git merge-base main HEAD) && git add -A && git commit -m "feat: ..."
3. Merge with the latest main: git merge origin/main
4. Force push to your branch: git push --force

We use pre-commit to help improve code quality. To initialize pre-commit, run:

pip install pre-commit
pre-commit install

You can also manually run pre-commit by:

pre-commit run

Adding CI tests

If possible, please add CI test(s) for your new feature:

1. Find the most relevant workflow yml file, which usually corresponds to a hydra default config (e.g. ppo_trainer, ppo_megatron_trainer, sft_trainer, etc).
2. Add related path patterns to the paths section if not already included.
3. Minimize the workload of the test script(s) (see existing scripts for examples).

---

We are HIRING! Send us an email if you are interested in internship/FTE opportunities in MLSys/LLM reasoning/multimodal alignment.