| Name | verl-cgft JSON |
| Version |
0.4.1.dev0
JSON |
| download |
| home_page | https://github.com/girishbarca/verl-fork |
| Summary | verl: Volcano Engine Reinforcement Learning for LLM |
| upload_time | 2025-07-27 17:30:46 |
| maintainer | None |
| docs_url | None |
| author | Bytedance - Seed - MLSys |
| requires_python | >=3.8 |
| license | Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
|
| keywords |
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
accelerate
codetiming
datasets
dill
flash-attn
hydra-core
liger-kernel
numpy
pandas
peft
pyarrow
pybind11
pylatexenc
pre-commit
ray
tensordict
torchdata
transformers
wandb
packaging
uvicorn
fastapi
latex2sympy2_extended
math_verify
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
<div align="center">
👋 Hi, everyone!
verl is a RL training library initiated by <b>ByteDance Seed team</b> and maintained by the verl community.
<br>
<br>
</div>
<div align="center">
[<img src="https://devin.ai/assets/deepwiki-badge.png" alt="Ask DeepWiki.com" height="20"/>](https://deepwiki.com/volcengine/verl)
[](https://github.com/volcengine/verl/stargazers)
[](https://twitter.com/verl_project)
<a href="https://join.slack.com/t/verlgroup/shared_invite/zt-2w5p9o4c3-yy0x2Q56s_VlGLsJ93A6vA"><img src="https://img.shields.io/badge/Slack-verl-blueviolet?logo=slack&"></a>
<a href="https://arxiv.org/pdf/2409.19256"><img src="https://img.shields.io/static/v1?label=EuroSys&message=Paper&color=red"></a>
[](https://verl.readthedocs.io/en/latest/)
<a href="https://raw.githubusercontent.com/eric-haibin-lin/verl-community/refs/heads/main/WeChat.JPG"><img src="https://img.shields.io/badge/微信-green?logo=wechat&"></a>
</div>
<h1 style="text-align: center;">verl: Volcano Engine Reinforcement Learning for LLMs</h1>
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](https://arxiv.org/abs/2409.19256v2)** 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.
</p>
## News
- [2025/06] verl with Megatron backend enables large MoE models such as [DeepSeek-671b and Qwen3-236b](https://verl.readthedocs.io/en/latest/perf/dpsk.html).
- [2025/06] verl team will provide latest project updates at [PyTorch Day China](https://www.lfasiallc.com/pytorch-day-china/) on June 7th. Meet our dev team in Beijing!
- [2025/05] [PF-PPO](https://arxiv.org/abs/2409.06957), accepted to ICML 2025, is now supported in verl! PF-PPO enhances policy learning efficiency and robustness by filtering potentially noisy reward signals and reusing high-quality experiences via a replay buffer.
- [2025/04] [Seed-Thinking-v1.5](https://github.com/ByteDance-Seed/Seed-Thinking-v1.5/blob/main/seed-thinking-v1.5.pdf) tech report is released! Trained with verl, Seed-Thinking-v1.5 achieves 86.7 on AIME 2024, 55.0 on Codeforces and 77.3 on GPQA, demonstrating excellent reasoning abilities in STEM and coding. Beyond reasoning tasks, the method demonstrates notable generalization across diverse domains.
- [2025/04] [VAPO](https://arxiv.org/pdf/2504.05118) (value-based augmented PPO) paper covers our latest RL method for reasoning models. Trained from Qwen-32B-base model, VAPO achieves 60.4 on AIME 2024, outperforming DAPO-32B.
- [2025/03] [DAPO](https://dapo-sia.github.io/) 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 available in `recipe/dapo` now.
<details><summary> more... </summary>
<ul>
<li>[2025/04] We will give a tutorial about latest post-training techniques and programming guide for verl at [ICLR 2025 Expo](https://iclr.cc/virtual/2025/calendar?filter_events=Expo+Talk+Panel&filter_rooms=), [SCI-FM workshop](https://open-foundation-model.github.io/) and [LMSys afterparty](https://lu.ma/d23nyynm). Talk materials available [here](https://github.com/eric-haibin-lin/verl-community/tree/main/iclr25). </li>
<li>[2025/03] verl v0.3.0.post1 is released! See [release note](https://github.com/volcengine/verl/releases/) for details. It achieves [~1.4x speedup](https://tongyx361.github.io/blogs/posts/verl-intro/#/verl-flexible-and-efficient-rl-for-llms) compared to prev versions.</li>
<li>[2025/05] verl will be presented at [A2M Shanghai](https://a2m.msup.com.cn/home/?aid=4488&city=shanghai) on 5/16 - 5/17.</li>
<li>[2025/05] verl will be presented at [GOSIM x PyTorch Day 2025](https://paris2025.gosim.org/). See you in Paris! </li>
<li>[2025/03] We introduced the programming model of verl at the [vLLM Beijing Meetup](https://mp.weixin.qq.com/s/n77GibL2corAtQHtVEAzfg) and [verl intro and updates](https://github.com/eric-haibin-lin/verl-community/blob/main/slides/verl-lmsys-meetup.pdf) at the [SGLang-LMSYS Org Meetup](https://lu.ma/ntjrr7ig) in Sunnyvale mid-March.</li>
<li>[2025/03] We will present verl(HybridFlow) at EuroSys 2025. See you in Rotterdam!</li>
<li>[2025/02] verl v0.2.0.post2 is released!</li>
<li>[2025/02] We presented verl in the <a href="https://lu.ma/ji7atxux">Bytedance/NVIDIA/Anyscale Ray Meetup</a>. See you in San Jose!</li>
<li>[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).</li>
<li>[2024/12] verl is presented at Ray Forward 2024. Slides available <a href="https://github.com/eric-haibin-lin/verl-community/blob/main/slides/Ray_Forward_2024_%E5%B7%AB%E9%94%A1%E6%96%8C.pdf">here</a></li>
<li>[2024/12] The team presented <a href="https://neurips.cc/Expo/Conferences/2024/workshop/100677">Post-training LLMs: From Algorithms to Infrastructure</a> at NeurIPS 2024. <a href="https://github.com/eric-haibin-lin/verl-data/tree/neurips">Slides</a> and <a href="https://neurips.cc/Expo/Conferences/2024/workshop/100677">video</a> available.</li>
<li>[2024/10] verl is presented at Ray Summit. <a href="https://www.youtube.com/watch?v=MrhMcXkXvJU&list=PLzTswPQNepXntmT8jr9WaNfqQ60QwW7-U&index=37">Youtube video</a> available.</li>
<li>[2024/08] HybridFlow (verl) is accepted to EuroSys 2025.</li>
</ul>
</details>
## Key Features
- **FSDP**, **FSDP2** and **Megatron-LM** for training.
- **vLLM**, **SGLang** and **HF Transformers** for rollout generation.
- Compatible with Hugging Face Transformers and Modelscope Hub: [Qwen-3](https://github.com/volcengine/verl/blob/main/examples/grpo_trainer/run_qwen3-8b.sh), Qwen-2.5, Llama3.1, Gemma2, DeepSeek-LLM, etc
- Supervised fine-tuning.
- Reinforcement learning with [PPO](examples/ppo_trainer/), [GRPO](examples/grpo_trainer/), [ReMax](examples/remax_trainer/), [REINFORCE++](https://verl.readthedocs.io/en/latest/examples/config.html#algorithm), [RLOO](examples/rloo_trainer/), [PRIME](recipe/prime/), [DAPO](recipe/dapo/), [DrGRPO](recipe/drgrpo), [KL_Cov & Clip_Cov](recipe/entropy) etc.
- Support model-based reward and function-based reward (verifiable reward) for math, [coding](https://github.com/volcengine/verl/tree/main/recipe/dapo), etc
- Support vision-language models (VLMs) and [multi-modal RL](examples/grpo_trainer/run_qwen2_5_vl-7b.sh) with Qwen2.5-vl, Kimi-VL
- [Multi-turn with tool calling](https://github.com/volcengine/verl/tree/main/examples/sglang_multiturn)
- LLM alignment recipes such as [Self-play preference optimization (SPPO)](https://github.com/volcengine/verl/tree/main/recipe/sppo)
- Flash attention 2, [sequence packing](examples/ppo_trainer/run_qwen2-7b_seq_balance.sh), [sequence parallelism](examples/ppo_trainer/run_deepseek7b_llm_sp2.sh) support via DeepSpeed Ulysses, [LoRA](examples/sft/gsm8k/run_qwen_05_peft.sh), [Liger-kernel](examples/sft/gsm8k/run_qwen_05_sp2_liger.sh).
- Scales up to 671B models and hundreds of GPUs with [expert parallelism](https://github.com/volcengine/verl/pull/1467)
- Multi-gpu [LoRA RL](https://verl.readthedocs.io/en/latest/advance/ppo_lora.html) support to save memory.
- Experiment tracking with wandb, swanlab, mlflow and tensorboard.
## Upcoming Features and Changes
- 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 and tools using optimizations https://github.com/volcengine/verl/issues/1882
- Environment interactions https://github.com/volcengine/verl/issues/1172
- List of breaking changes since v0.3 https://github.com/volcengine/verl/discussions/943, entropy_coeff defaults to 0
- Lora for RL https://github.com/volcengine/verl/pull/1127
## Getting Started
<a href="https://verl.readthedocs.io/en/latest/index.html"><b>Documentation</b></a>
**Quickstart:**
- [Installation](https://verl.readthedocs.io/en/latest/start/install.html)
- [Quickstart](https://verl.readthedocs.io/en/latest/start/quickstart.html)
- [Programming Guide](https://verl.readthedocs.io/en/latest/hybrid_flow.html) & [Tech Talk](https://hcqnc.xetlk.com/sl/3vACOK) (in Chinese)
- [PPO in verl](https://verl.readthedocs.io/en/latest/algo/ppo.html)
- [GRPO in verl](https://verl.readthedocs.io/en/latest/algo/grpo.html)
**Running a PPO example step-by-step:**
- [Prepare Data for Post-Training](https://verl.readthedocs.io/en/latest/preparation/prepare_data.html)
- [Implement Reward Function for Dataset](https://verl.readthedocs.io/en/latest/preparation/reward_function.html)
- [PPO Example Architecture](https://verl.readthedocs.io/en/latest/examples/ppo_code_architecture.html)
- [Config Explanation](https://verl.readthedocs.io/en/latest/examples/config.html)
**Reproducible algorithm baselines:**
- [RL performance on coding, math](https://verl.readthedocs.io/en/latest/algo/baseline.html)
**For code explanation and advance usage (extension):**
- PPO Trainer and Workers
- [PPO Ray Trainer](https://verl.readthedocs.io/en/latest/workers/ray_trainer.html)
- [PyTorch FSDP Backend](https://verl.readthedocs.io/en/latest/workers/fsdp_workers.html)
- [Megatron-LM Backend](https://verl.readthedocs.io/en/latest/index.html)
- Advanced Usage and Extension
- [Add Models with the FSDP Backend](https://verl.readthedocs.io/en/latest/advance/fsdp_extension.html)
- [Add Models with the Megatron-LM Backend](https://verl.readthedocs.io/en/latest/advance/megatron_extension.html)
- [Multi-turn Rollout Support](https://verl.readthedocs.io/en/latest/sglang_multiturn/multiturn.html)
- [Search Tool Integration](https://verl.readthedocs.io/en/latest/sglang_multiturn/search_tool_example.html)
- [Sandbox Fusion Integration](https://verl.readthedocs.io/en/latest/examples/sandbox_fusion_example.html)
- [Deployment using Separate GPU Resources](https://github.com/volcengine/verl/tree/main/examples/split_placement)
- [Extend to Other RL(HF) algorithms](https://verl.readthedocs.io/en/latest/advance/dpo_extension.html)
- [Ray API design tutorial](https://verl.readthedocs.io/en/latest/advance/placement.html)
**Blogs from the community**
- [When Reasoning Models Break Tokenization: The Hidden Complexity of Multiturn Training](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/fast_tokenization/multiturn_tokenization_and_masking.md)
- [verl deployment on AWS SageMaker](https://medium.com/@kaige.yang0110/run-verl-on-sagemaker-using-4x8-l40s-gpus-8e6d5c3c61d3)
- [verl x SGLang Multi-turn Code Walkthrough](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/code-walk-through/readme_EN.md)
- [Optimizing SGLang Memory Usage in verl](https://hebiao064.github.io/rl-memory-management)
- [SGLang, verl, OpenBMB and Tsinghua University: Pioneering End-to-End Multi-Turn RLHF](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/verl-multiturn-rollout-Release.md)
- [Reinforcement Learning from Human Feedback on AMD GPUs with verl and ROCm Integration](https://rocm.blogs.amd.com/artificial-intelligence/verl-large-scale/README.html)
- [veMLP x verl ：玩转强化学习训练](https://mp.weixin.qq.com/s/7nbqxk4knMGd-hQE9ls2tA)
- [使用 verl 进行 GRPO 分布式强化学习训练最佳实践](https://www.volcengine.com/docs/6459/1463942)
- [HybridFlow verl 原文浅析](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/readme.md)
- [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！](https://team.doubao.com/en/blog/%E6%9C%80%E9%AB%98%E6%8F%90%E5%8D%8720%E5%80%8D%E5%90%9E%E5%90%90%E9%87%8F-%E8%B1%86%E5%8C%85%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%9B%A2%E9%98%9F%E5%8F%91%E5%B8%83%E5%85%A8%E6%96%B0-rlhf-%E6%A1%86%E6%9E%B6-%E7%8E%B0%E5%B7%B2%E5%BC%80%E6%BA%90)
## Performance Tuning Guide
The performance is essential for on-policy RL algorithm. We have written a detailed [performance tuning guide](https://verl.readthedocs.io/en/latest/perf/perf_tuning.html) 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](https://github.com/volcengine/verl/blob/main/docs/README_vllm0.8.md) 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](https://verl.readthedocs.io/en/latest/workers/sglang_worker.html) for the installation guide and more information.
## Upgrade to FSDP2
verl is fully embracing FSDP2! FSDP2 is recommended by torch distributed team, providing better throughput and memory usage, and is composible with other features (e.g. torch.compile). To enable FSDP2, simply use verl main and set the following options:
```
actor_rollout_ref.ref.strategy=fsdp2
actor_rollout_ref.actor.strategy=fsdp2
critic.strategy=fsdp2
reward_model.strategy=fsdp2
```
Furthermore, FSDP2 cpu offloading is compatible with gradient accumulation. You can turn it on to save memory with `actor_rollout_ref.actor.fsdp_config.offload_policy=True`. For more details, see https://github.com/volcengine/verl/pull/1026
## AMD Support (ROCm Kernel)
verl now supports FSDP as the training engine (Megatron support coming soon) and both integrates with vLLM and SGLang as inference engines. Please refer to [this document](https://github.com/volcengine/verl/blob/main/docs/amd_tutorial/amd_build_dockerfile_page.rst) for the installation guide and more information, and [this document](https://github.com/volcengine/verl/blob/main/docs/amd_tutorial/amd_vllm_page.rst) for the vLLM performance tuning for ROCm.
## Citation and acknowledgement
If you find the project helpful, please cite:
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://arxiv.org/abs/2409.19256v2)
- [A Framework for Training Large Language Models for Code Generation via Proximal Policy Optimization](https://i.cs.hku.hk/~cwu/papers/gmsheng-NL2Code24.pdf)
```bibtex
@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](https://github.com/QwenLM/), Shanghai AI Lab, Tsinghua University, UC Berkeley, UCLA, UIUC, University of Hong Kong, ke.com, [All Hands AI](https://www.all-hands.dev/), [ModelBest](http://modelbest.cn/), JD AI Lab, Microsoft Research, [StepFun](https://www.stepfun.com/), Amazon, LinkedIn, Meituan, [Camel-AI](https://www.camel-ai.org/), [OpenManus](https://github.com/OpenManus), Xiaomi, NVIDIA research, [Baichuan](https://www.baichuan-ai.com/home), [RedNote](https://www.xiaohongshu.com/), [SwissAI](https://www.swiss-ai.org/), [Moonshot AI (Kimi)](https://www.moonshot-ai.com/), Baidu, Snowflake, Skywork.ai, JetBrains, [IceSword Lab](https://www.iceswordlab.com), and many more.
## Awesome work using verl
- [TinyZero](https://github.com/Jiayi-Pan/TinyZero): a reproduction of **DeepSeek R1 Zero** recipe for reasoning tasks 
- [SkyThought](https://github.com/NovaSky-AI/SkyThought): RL training for Sky-T1-7B by NovaSky AI team. 
- [simpleRL-reason](https://github.com/hkust-nlp/simpleRL-reason): SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild 
- [Easy-R1](https://github.com/hiyouga/EasyR1): **Multi-modal** RL training framework 
- [OpenManus-RL](https://github.com/OpenManus/OpenManus-RL): LLM Agents RL tunning framework for multiple agent environments. 
- [rllm](https://github.com/agentica-project/rllm): async RL training with [verl-pipeline](https://github.com/agentica-project/verl-pipeline) 
- [RAGEN](https://github.com/ZihanWang314/ragen): a general-purpose reasoning **agent** training framework 
- [Search-R1](https://github.com/PeterGriffinJin/Search-R1): RL with reasoning and **searching (tool-call)** interleaved LLMs 
- [ReSearch](https://github.com/Agent-RL/ReSearch): Learning to **Re**ason with **Search** for LLMs via Reinforcement Learning 
- [Skywork-OR1](https://github.com/SkyworkAI/Skywork-OR1): Skywork open reaonser series 
- [ToRL](https://github.com/GAIR-NLP/ToRL): Scaling tool-integrated RL 
- [Absolute Zero Reasoner](https://github.com/LeapLabTHU/Absolute-Zero-Reasoner): [A no human curated data self-play framework for reasoning](https://arxiv.org/abs/2505.03335) 
- [verl-agent](https://github.com/langfengQ/verl-agent): A scalable training framework for **long-horizon LLM/VLM agents**, along with a new algorithm **GiGPO** 
- [RL-Factory](https://github.com/Simple-Efficient/RL-Factory): An easy and efficient RL post-training framework for Agentic Learning 
- [ReTool](https://retool-rl.github.io/): ReTool: reinforcement learning for strategic tool use in LLMs. Code release is in progress...
- [verl-tool](https://github.com/TIGER-AI-Lab/verl-tool): An unified and easy-to-extend tool-agent training framework based on verl
- [PRIME](https://github.com/PRIME-RL/PRIME): Process reinforcement through implicit rewards 
- [GUI-R1](https://github.com/ritzz-ai/GUI-R1): **GUI-R1**: A Generalist R1-style Vision-Language Action Model For **GUI Agents** 
- [DeepRetrieval](https://github.com/pat-jj/DeepRetrieval): RL Training of **Search Agent** with **Search/Retrieval Outcome** 
- [Code-R1](https://github.com/ganler/code-r1): Reproducing R1 for **Code** with Reliable Rewards 
- [DeepResearcher](https://github.com/GAIR-NLP/DeepResearcher): Scaling deep research via reinforcement learning in real-world environments 
- [VAGEN](https://github.com/RAGEN-AI/VAGEN): Training VLM agents with multi-turn reinforcement learning 
- [RM-R1](https://arxiv.org/abs/2505.02387): RL training of reasoning reward models 
- [LUFFY](https://arxiv.org/pdf/2504.14945): Learning to Reason under Off-Policy Guidance
- [DeepMath](https://github.com/zwhe99/DeepMath): DeepMath-103K data and series models for math reasoning
- [Entropy Mechanism of RL](https://github.com/PRIME-RL/Entropy-Mechanism-of-RL): The Entropy Mechanism of Reinforcement Learning for Large Language Model Reasoning
- [LLaSA-TTS-GRPO](https://github.com/channel-io/ch-tts-llasa-rl-grpo): TTS fine-tuning with GRPO optimization based on LLASA models 
- [PF-PPO](https://arxiv.org/abs/2409.06957): Policy Filtration for PPO based on the reliability of reward signals for more efficient and robust RLHF.
- [RACRO](https://github.com/gyhdog99/RACRO2): Build multi-modal reasoning models via decoupling it into query-conditioned captioning and text-only reasoning 
and many more awesome work listed in [recipe](recipe/README.md).
## Contribution Guide
See [contributions guide](CONTRIBUTING.md)
## About [ByteDance Seed Team](https://team.doubao.com/)
Founded in 2023, ByteDance Seed Team is dedicated to crafting the industry's most advanced AI foundation models. The team aspires to become a world-class research team and make significant contributions to the advancement of science and society. You can get to know Bytedance Seed better through the following channels👇
<div>
<a href="https://team.doubao.com/">
<img src="https://img.shields.io/badge/Website-%231e37ff?style=for-the-badge&logo=bytedance&logoColor=white"></a>
<a href="https://github.com/user-attachments/assets/469535a8-42f2-4797-acdf-4f7a1d4a0c3e">
<img src="https://img.shields.io/badge/WeChat-07C160?style=for-the-badge&logo=wechat&logoColor=white"></a>
<a href="https://www.xiaohongshu.com/user/profile/668e7e15000000000303157d?xsec_token=ABl2-aqekpytY6A8TuxjrwnZskU-6BsMRE_ufQQaSAvjc%3D&xsec_source=pc_search">
<img src="https://img.shields.io/badge/Xiaohongshu-%23FF2442?style=for-the-badge&logo=xiaohongshu&logoColor=white"></a>
<a href="https://www.zhihu.com/org/dou-bao-da-mo-xing-tuan-dui/">
<img src="https://img.shields.io/badge/zhihu-%230084FF?style=for-the-badge&logo=zhihu&logoColor=white"></a>
</div>
---
We are HIRING! Send us an [email](mailto:haibin.lin@bytedance.com) if you are interested in internship/FTE opportunities in RL for agents.
Raw data
{
"_id": null,
"home_page": "https://github.com/girishbarca/verl-fork",
"name": "verl-cgft",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.8",
"maintainer_email": null,
"keywords": null,
"author": "Bytedance - Seed - MLSys",
"author_email": "zhangchi.usc1992@bytedance.com, gmsheng@connect.hku.hk",
"download_url": "https://files.pythonhosted.org/packages/26/7a/6d38c181978c2d5026262cf35ccc0602f049be07fc19c97ae1829c1d0ecd/verl_cgft-0.4.1.dev0.tar.gz",
"platform": null,
"description": "<div align=\"center\">\n \ud83d\udc4b Hi, everyone! \n verl is a RL training library initiated by <b>ByteDance Seed team</b> and maintained by the verl community.\n <br>\n <br>\n</div>\n\n<div align=\"center\">\n\n[<img src=\"https://devin.ai/assets/deepwiki-badge.png\" alt=\"Ask DeepWiki.com\" height=\"20\"/>](https://deepwiki.com/volcengine/verl)\n[](https://github.com/volcengine/verl/stargazers)\n[](https://twitter.com/verl_project)\n<a href=\"https://join.slack.com/t/verlgroup/shared_invite/zt-2w5p9o4c3-yy0x2Q56s_VlGLsJ93A6vA\"><img src=\"https://img.shields.io/badge/Slack-verl-blueviolet?logo=slack&\"></a>\n<a href=\"https://arxiv.org/pdf/2409.19256\"><img src=\"https://img.shields.io/static/v1?label=EuroSys&message=Paper&color=red\"></a>\n[](https://verl.readthedocs.io/en/latest/)\n<a href=\"https://raw.githubusercontent.com/eric-haibin-lin/verl-community/refs/heads/main/WeChat.JPG\"><img src=\"https://img.shields.io/badge/\u5fae\u4fe1-green?logo=wechat&\"></a>\n\n</div>\n\n\n\n<h1 style=\"text-align: center;\">verl: Volcano Engine Reinforcement Learning for LLMs</h1>\n\nverl is a flexible, efficient and production-ready RL training library for large language models (LLMs).\n\nverl is the open-source version of **[HybridFlow: A Flexible and Efficient RLHF Framework](https://arxiv.org/abs/2409.19256v2)** paper.\n\nverl is flexible and easy to use with:\n\n- **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.\n\n- **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\n\n- **Flexible device mapping**: Supports various placement of models onto different sets of GPUs for efficient resource utilization and scalability across different cluster sizes.\n\n- Ready integration with popular HuggingFace models\n\nverl is fast with:\n\n- **State-of-the-art throughput**: SOTA LLM training and inference engine integrations and SOTA RL throughput.\n\n- **Efficient actor model resharding with 3D-HybridEngine**: Eliminates memory redundancy and significantly reduces communication overhead during transitions between training and generation phases.\n\n</p>\n\n## News\n\n- [2025/06] verl with Megatron backend enables large MoE models such as [DeepSeek-671b and Qwen3-236b](https://verl.readthedocs.io/en/latest/perf/dpsk.html).\n- [2025/06] verl team will provide latest project updates at [PyTorch Day China](https://www.lfasiallc.com/pytorch-day-china/) on June 7th. Meet our dev team in Beijing!\n- [2025/05] [PF-PPO](https://arxiv.org/abs/2409.06957), accepted to ICML 2025, is now supported in verl! PF-PPO enhances policy learning efficiency and robustness by filtering potentially noisy reward signals and reusing high-quality experiences via a replay buffer.\n- [2025/04] [Seed-Thinking-v1.5](https://github.com/ByteDance-Seed/Seed-Thinking-v1.5/blob/main/seed-thinking-v1.5.pdf) tech report is released! Trained with verl, Seed-Thinking-v1.5 achieves 86.7 on AIME 2024, 55.0 on Codeforces and 77.3 on GPQA, demonstrating excellent reasoning abilities in STEM and coding. Beyond reasoning tasks, the method demonstrates notable generalization across diverse domains.\n- [2025/04] [VAPO](https://arxiv.org/pdf/2504.05118) (value-based augmented PPO) paper covers our latest RL method for reasoning models. Trained from Qwen-32B-base model, VAPO achieves 60.4 on AIME 2024, outperforming DAPO-32B.\n- [2025/03] [DAPO](https://dapo-sia.github.io/) 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 available in `recipe/dapo` now.\n<details><summary> more... </summary>\n<ul>\n\n <li>[2025/04] We will give a tutorial about latest post-training techniques and programming guide for verl at [ICLR 2025 Expo](https://iclr.cc/virtual/2025/calendar?filter_events=Expo+Talk+Panel&filter_rooms=), [SCI-FM workshop](https://open-foundation-model.github.io/) and [LMSys afterparty](https://lu.ma/d23nyynm). Talk materials available [here](https://github.com/eric-haibin-lin/verl-community/tree/main/iclr25). </li>\n <li>[2025/03] verl v0.3.0.post1 is released! See [release note](https://github.com/volcengine/verl/releases/) for details. It achieves [~1.4x speedup](https://tongyx361.github.io/blogs/posts/verl-intro/#/verl-flexible-and-efficient-rl-for-llms) compared to prev versions.</li>\n <li>[2025/05] verl will be presented at [A2M Shanghai](https://a2m.msup.com.cn/home/?aid=4488&city=shanghai) on 5/16 - 5/17.</li>\n <li>[2025/05] verl will be presented at [GOSIM x PyTorch Day 2025](https://paris2025.gosim.org/). See you in Paris! </li>\n <li>[2025/03] We introduced the programming model of verl at the [vLLM Beijing Meetup](https://mp.weixin.qq.com/s/n77GibL2corAtQHtVEAzfg) and [verl intro and updates](https://github.com/eric-haibin-lin/verl-community/blob/main/slides/verl-lmsys-meetup.pdf) at the [SGLang-LMSYS Org Meetup](https://lu.ma/ntjrr7ig) in Sunnyvale mid-March.</li>\n <li>[2025/03] We will present verl(HybridFlow) at EuroSys 2025. See you in Rotterdam!</li>\n <li>[2025/02] verl v0.2.0.post2 is released!</li>\n <li>[2025/02] We presented verl in the <a href=\"https://lu.ma/ji7atxux\">Bytedance/NVIDIA/Anyscale Ray Meetup</a>. See you in San Jose!</li>\n <li>[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).</li>\n <li>[2024/12] verl is presented at Ray Forward 2024. Slides available <a href=\"https://github.com/eric-haibin-lin/verl-community/blob/main/slides/Ray_Forward_2024_%E5%B7%AB%E9%94%A1%E6%96%8C.pdf\">here</a></li>\n <li>[2024/12] The team presented <a href=\"https://neurips.cc/Expo/Conferences/2024/workshop/100677\">Post-training LLMs: From Algorithms to Infrastructure</a> at NeurIPS 2024. <a href=\"https://github.com/eric-haibin-lin/verl-data/tree/neurips\">Slides</a> and <a href=\"https://neurips.cc/Expo/Conferences/2024/workshop/100677\">video</a> available.</li>\n <li>[2024/10] verl is presented at Ray Summit. <a href=\"https://www.youtube.com/watch?v=MrhMcXkXvJU&list=PLzTswPQNepXntmT8jr9WaNfqQ60QwW7-U&index=37\">Youtube video</a> available.</li>\n <li>[2024/08] HybridFlow (verl) is accepted to EuroSys 2025.</li>\n</ul> \n</details>\n\n## Key Features\n\n- **FSDP**, **FSDP2** and **Megatron-LM** for training.\n- **vLLM**, **SGLang** and **HF Transformers** for rollout generation.\n- Compatible with Hugging Face Transformers and Modelscope Hub: [Qwen-3](https://github.com/volcengine/verl/blob/main/examples/grpo_trainer/run_qwen3-8b.sh), Qwen-2.5, Llama3.1, Gemma2, DeepSeek-LLM, etc\n- Supervised fine-tuning.\n- Reinforcement learning with [PPO](examples/ppo_trainer/), [GRPO](examples/grpo_trainer/), [ReMax](examples/remax_trainer/), [REINFORCE++](https://verl.readthedocs.io/en/latest/examples/config.html#algorithm), [RLOO](examples/rloo_trainer/), [PRIME](recipe/prime/), [DAPO](recipe/dapo/), [DrGRPO](recipe/drgrpo), [KL_Cov & Clip_Cov](recipe/entropy) etc.\n - Support model-based reward and function-based reward (verifiable reward) for math, [coding](https://github.com/volcengine/verl/tree/main/recipe/dapo), etc\n - Support vision-language models (VLMs) and [multi-modal RL](examples/grpo_trainer/run_qwen2_5_vl-7b.sh) with Qwen2.5-vl, Kimi-VL\n - [Multi-turn with tool calling](https://github.com/volcengine/verl/tree/main/examples/sglang_multiturn)\n- LLM alignment recipes such as [Self-play preference optimization (SPPO)](https://github.com/volcengine/verl/tree/main/recipe/sppo)\n- Flash attention 2, [sequence packing](examples/ppo_trainer/run_qwen2-7b_seq_balance.sh), [sequence parallelism](examples/ppo_trainer/run_deepseek7b_llm_sp2.sh) support via DeepSpeed Ulysses, [LoRA](examples/sft/gsm8k/run_qwen_05_peft.sh), [Liger-kernel](examples/sft/gsm8k/run_qwen_05_sp2_liger.sh).\n- Scales up to 671B models and hundreds of GPUs with [expert parallelism](https://github.com/volcengine/verl/pull/1467)\n- Multi-gpu [LoRA RL](https://verl.readthedocs.io/en/latest/advance/ppo_lora.html) support to save memory.\n- Experiment tracking with wandb, swanlab, mlflow and tensorboard.\n\n## Upcoming Features and Changes\n\n- Roadmap https://github.com/volcengine/verl/issues/710\n- DeepSeek 671b optimizations with Megatron v0.11 https://github.com/volcengine/verl/issues/708\n- Multi-turn rollout and tools using optimizations https://github.com/volcengine/verl/issues/1882\n- Environment interactions https://github.com/volcengine/verl/issues/1172\n- List of breaking changes since v0.3 https://github.com/volcengine/verl/discussions/943, entropy_coeff defaults to 0\n- Lora for RL https://github.com/volcengine/verl/pull/1127 \n\n## Getting Started\n\n<a href=\"https://verl.readthedocs.io/en/latest/index.html\"><b>Documentation</b></a>\n\n**Quickstart:**\n\n- [Installation](https://verl.readthedocs.io/en/latest/start/install.html)\n- [Quickstart](https://verl.readthedocs.io/en/latest/start/quickstart.html)\n- [Programming Guide](https://verl.readthedocs.io/en/latest/hybrid_flow.html) & [Tech Talk](https://hcqnc.xetlk.com/sl/3vACOK) (in Chinese)\n- [PPO in verl](https://verl.readthedocs.io/en/latest/algo/ppo.html)\n- [GRPO in verl](https://verl.readthedocs.io/en/latest/algo/grpo.html)\n\n**Running a PPO example step-by-step:**\n\n\n- [Prepare Data for Post-Training](https://verl.readthedocs.io/en/latest/preparation/prepare_data.html)\n- [Implement Reward Function for Dataset](https://verl.readthedocs.io/en/latest/preparation/reward_function.html)\n- [PPO Example Architecture](https://verl.readthedocs.io/en/latest/examples/ppo_code_architecture.html)\n- [Config Explanation](https://verl.readthedocs.io/en/latest/examples/config.html)\n\n**Reproducible algorithm baselines:**\n\n- [RL performance on coding, math](https://verl.readthedocs.io/en/latest/algo/baseline.html)\n\n**For code explanation and advance usage (extension):**\n\n- PPO Trainer and Workers\n - [PPO Ray Trainer](https://verl.readthedocs.io/en/latest/workers/ray_trainer.html)\n - [PyTorch FSDP Backend](https://verl.readthedocs.io/en/latest/workers/fsdp_workers.html)\n - [Megatron-LM Backend](https://verl.readthedocs.io/en/latest/index.html)\n\n- Advanced Usage and Extension\n - [Add Models with the FSDP Backend](https://verl.readthedocs.io/en/latest/advance/fsdp_extension.html)\n - [Add Models with the Megatron-LM Backend](https://verl.readthedocs.io/en/latest/advance/megatron_extension.html)\n - [Multi-turn Rollout Support](https://verl.readthedocs.io/en/latest/sglang_multiturn/multiturn.html)\n - [Search Tool Integration](https://verl.readthedocs.io/en/latest/sglang_multiturn/search_tool_example.html)\n - [Sandbox Fusion Integration](https://verl.readthedocs.io/en/latest/examples/sandbox_fusion_example.html)\n - [Deployment using Separate GPU Resources](https://github.com/volcengine/verl/tree/main/examples/split_placement)\n - [Extend to Other RL(HF) algorithms](https://verl.readthedocs.io/en/latest/advance/dpo_extension.html)\n - [Ray API design tutorial](https://verl.readthedocs.io/en/latest/advance/placement.html)\n\n**Blogs from the community**\n\n- [When Reasoning Models Break Tokenization: The Hidden Complexity of Multiturn Training](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/fast_tokenization/multiturn_tokenization_and_masking.md)\n- [verl deployment on AWS SageMaker](https://medium.com/@kaige.yang0110/run-verl-on-sagemaker-using-4x8-l40s-gpus-8e6d5c3c61d3)\n- [verl x SGLang Multi-turn Code Walkthrough](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/code-walk-through/readme_EN.md)\n- [Optimizing SGLang Memory Usage in verl](https://hebiao064.github.io/rl-memory-management)\n- [SGLang, verl, OpenBMB and Tsinghua University: Pioneering End-to-End Multi-Turn RLHF](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/multi-turn/verl-multiturn-rollout-Release.md)\n- [Reinforcement Learning from Human Feedback on AMD GPUs with verl and ROCm Integration](https://rocm.blogs.amd.com/artificial-intelligence/verl-large-scale/README.html)\n- [veMLP x verl \uff1a\u73a9\u8f6c\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3](https://mp.weixin.qq.com/s/7nbqxk4knMGd-hQE9ls2tA)\n- [\u4f7f\u7528 verl \u8fdb\u884c GRPO \u5206\u5e03\u5f0f\u5f3a\u5316\u5b66\u4e60\u8bad\u7ec3\u6700\u4f73\u5b9e\u8df5](https://www.volcengine.com/docs/6459/1463942)\n- [HybridFlow verl \u539f\u6587\u6d45\u6790](https://github.com/zhaochenyang20/Awesome-ML-SYS-Tutorial/blob/main/rlhf/verl/readme.md)\n- [\u6700\u9ad8\u63d0\u5347 20 \u500d\u541e\u5410\u91cf\uff01\u8c46\u5305\u5927\u6a21\u578b\u56e2\u961f\u53d1\u5e03\u5168\u65b0 RLHF \u6846\u67b6\uff0c\u73b0\u5df2\u5f00\u6e90\uff01](https://team.doubao.com/en/blog/%E6%9C%80%E9%AB%98%E6%8F%90%E5%8D%8720%E5%80%8D%E5%90%9E%E5%90%90%E9%87%8F-%E8%B1%86%E5%8C%85%E5%A4%A7%E6%A8%A1%E5%9E%8B%E5%9B%A2%E9%98%9F%E5%8F%91%E5%B8%83%E5%85%A8%E6%96%B0-rlhf-%E6%A1%86%E6%9E%B6-%E7%8E%B0%E5%B7%B2%E5%BC%80%E6%BA%90)\n\n## Performance Tuning Guide\n\nThe performance is essential for on-policy RL algorithm. We have written a detailed [performance tuning guide](https://verl.readthedocs.io/en/latest/perf/perf_tuning.html) to help you optimize performance.\n\n## Upgrade to vLLM >= v0.8.2\n\nverl now supports vLLM>=0.8.2 when using FSDP as the training backend. Please refer to [this document](https://github.com/volcengine/verl/blob/main/docs/README_vllm0.8.md) for the installation guide and more information. Please avoid vllm 0.7.x, which contains bugs that may lead to OOMs and unexpected errors.\n\n## Use Latest SGLang\n\nSGLang 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](https://verl.readthedocs.io/en/latest/workers/sglang_worker.html) for the installation guide and more information.\n\n## Upgrade to FSDP2\n\nverl is fully embracing FSDP2! FSDP2 is recommended by torch distributed team, providing better throughput and memory usage, and is composible with other features (e.g. torch.compile). To enable FSDP2, simply use verl main and set the following options:\n```\nactor_rollout_ref.ref.strategy=fsdp2\nactor_rollout_ref.actor.strategy=fsdp2\ncritic.strategy=fsdp2 \nreward_model.strategy=fsdp2 \n```\nFurthermore, FSDP2 cpu offloading is compatible with gradient accumulation. You can turn it on to save memory with `actor_rollout_ref.actor.fsdp_config.offload_policy=True`. For more details, see https://github.com/volcengine/verl/pull/1026\n\n## AMD Support (ROCm Kernel)\n\nverl now supports FSDP as the training engine (Megatron support coming soon) and both integrates with vLLM and SGLang as inference engines. Please refer to [this document](https://github.com/volcengine/verl/blob/main/docs/amd_tutorial/amd_build_dockerfile_page.rst) for the installation guide and more information, and [this document](https://github.com/volcengine/verl/blob/main/docs/amd_tutorial/amd_vllm_page.rst) for the vLLM performance tuning for ROCm.\n\n\n## Citation and acknowledgement\n\nIf you find the project helpful, please cite:\n\n- [HybridFlow: A Flexible and Efficient RLHF Framework](https://arxiv.org/abs/2409.19256v2)\n- [A Framework for Training Large Language Models for Code Generation via Proximal Policy Optimization](https://i.cs.hku.hk/~cwu/papers/gmsheng-NL2Code24.pdf)\n\n```bibtex\n@article{sheng2024hybridflow,\n title = {HybridFlow: A Flexible and Efficient RLHF Framework},\n 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},\n year = {2024},\n journal = {arXiv preprint arXiv: 2409.19256}\n}\n```\n\nverl 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](https://github.com/QwenLM/), Shanghai AI Lab, Tsinghua University, UC Berkeley, UCLA, UIUC, University of Hong Kong, ke.com, [All Hands AI](https://www.all-hands.dev/), [ModelBest](http://modelbest.cn/), JD AI Lab, Microsoft Research, [StepFun](https://www.stepfun.com/), Amazon, LinkedIn, Meituan, [Camel-AI](https://www.camel-ai.org/), [OpenManus](https://github.com/OpenManus), Xiaomi, NVIDIA research, [Baichuan](https://www.baichuan-ai.com/home), [RedNote](https://www.xiaohongshu.com/), [SwissAI](https://www.swiss-ai.org/), [Moonshot AI (Kimi)](https://www.moonshot-ai.com/), Baidu, Snowflake, Skywork.ai, JetBrains, [IceSword Lab](https://www.iceswordlab.com), and many more.\n\n## Awesome work using verl\n\n- [TinyZero](https://github.com/Jiayi-Pan/TinyZero): a reproduction of **DeepSeek R1 Zero** recipe for reasoning tasks \n- [SkyThought](https://github.com/NovaSky-AI/SkyThought): RL training for Sky-T1-7B by NovaSky AI team. \n- [simpleRL-reason](https://github.com/hkust-nlp/simpleRL-reason): SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wild \n- [Easy-R1](https://github.com/hiyouga/EasyR1): **Multi-modal** RL training framework \n- [OpenManus-RL](https://github.com/OpenManus/OpenManus-RL): LLM Agents RL tunning framework for multiple agent environments. \n- [rllm](https://github.com/agentica-project/rllm): async RL training with [verl-pipeline](https://github.com/agentica-project/verl-pipeline) \n- [RAGEN](https://github.com/ZihanWang314/ragen): a general-purpose reasoning **agent** training framework \n- [Search-R1](https://github.com/PeterGriffinJin/Search-R1): RL with reasoning and **searching (tool-call)** interleaved LLMs \n- [ReSearch](https://github.com/Agent-RL/ReSearch): Learning to **Re**ason with **Search** for LLMs via Reinforcement Learning \n- [Skywork-OR1](https://github.com/SkyworkAI/Skywork-OR1): Skywork open reaonser series \n- [ToRL](https://github.com/GAIR-NLP/ToRL): Scaling tool-integrated RL \n- [Absolute Zero Reasoner](https://github.com/LeapLabTHU/Absolute-Zero-Reasoner): [A no human curated data self-play framework for reasoning](https://arxiv.org/abs/2505.03335) \n- [verl-agent](https://github.com/langfengQ/verl-agent): A scalable training framework for **long-horizon LLM/VLM agents**, along with a new algorithm **GiGPO** \n- [RL-Factory](https://github.com/Simple-Efficient/RL-Factory): An easy and efficient RL post-training framework for Agentic Learning \n- [ReTool](https://retool-rl.github.io/): ReTool: reinforcement learning for strategic tool use in LLMs. Code release is in progress...\n- [verl-tool](https://github.com/TIGER-AI-Lab/verl-tool): An unified and easy-to-extend tool-agent training framework based on verl\n- [PRIME](https://github.com/PRIME-RL/PRIME): Process reinforcement through implicit rewards \n- [GUI-R1](https://github.com/ritzz-ai/GUI-R1): **GUI-R1**: A Generalist R1-style Vision-Language Action Model For **GUI Agents** \n- [DeepRetrieval](https://github.com/pat-jj/DeepRetrieval): RL Training of **Search Agent** with **Search/Retrieval Outcome** \n- [Code-R1](https://github.com/ganler/code-r1): Reproducing R1 for **Code** with Reliable Rewards \n- [DeepResearcher](https://github.com/GAIR-NLP/DeepResearcher): Scaling deep research via reinforcement learning in real-world environments \n- [VAGEN](https://github.com/RAGEN-AI/VAGEN): Training VLM agents with multi-turn reinforcement learning \n- [RM-R1](https://arxiv.org/abs/2505.02387): RL training of reasoning reward models \n- [LUFFY](https://arxiv.org/pdf/2504.14945): Learning to Reason under Off-Policy Guidance\n- [DeepMath](https://github.com/zwhe99/DeepMath): DeepMath-103K data and series models for math reasoning\n- [Entropy Mechanism of RL](https://github.com/PRIME-RL/Entropy-Mechanism-of-RL): The Entropy Mechanism of Reinforcement Learning for Large Language Model Reasoning\n- [LLaSA-TTS-GRPO](https://github.com/channel-io/ch-tts-llasa-rl-grpo): TTS fine-tuning with GRPO optimization based on LLASA models \n- [PF-PPO](https://arxiv.org/abs/2409.06957): Policy Filtration for PPO based on the reliability of reward signals for more efficient and robust RLHF.\n- [RACRO](https://github.com/gyhdog99/RACRO2): Build multi-modal reasoning models via decoupling it into query-conditioned captioning and text-only reasoning \n\nand many more awesome work listed in [recipe](recipe/README.md).\n\n## Contribution Guide\n\nSee [contributions guide](CONTRIBUTING.md)\n\n## About [ByteDance Seed Team](https://team.doubao.com/)\n\nFounded in 2023, ByteDance Seed Team is dedicated to crafting the industry's most advanced AI foundation models. The team aspires to become a world-class research team and make significant contributions to the advancement of science and society. You can get to know Bytedance Seed better through the following channels\ud83d\udc47\n<div>\n <a href=\"https://team.doubao.com/\">\n <img src=\"https://img.shields.io/badge/Website-%231e37ff?style=for-the-badge&logo=bytedance&logoColor=white\"></a>\n <a href=\"https://github.com/user-attachments/assets/469535a8-42f2-4797-acdf-4f7a1d4a0c3e\">\n <img src=\"https://img.shields.io/badge/WeChat-07C160?style=for-the-badge&logo=wechat&logoColor=white\"></a>\n <a href=\"https://www.xiaohongshu.com/user/profile/668e7e15000000000303157d?xsec_token=ABl2-aqekpytY6A8TuxjrwnZskU-6BsMRE_ufQQaSAvjc%3D&xsec_source=pc_search\">\n <img src=\"https://img.shields.io/badge/Xiaohongshu-%23FF2442?style=for-the-badge&logo=xiaohongshu&logoColor=white\"></a>\n <a href=\"https://www.zhihu.com/org/dou-bao-da-mo-xing-tuan-dui/\">\n <img src=\"https://img.shields.io/badge/zhihu-%230084FF?style=for-the-badge&logo=zhihu&logoColor=white\"></a>\n\n</div>\n---\n\nWe are HIRING! Send us an [email](mailto:haibin.lin@bytedance.com) if you are interested in internship/FTE opportunities in RL for agents.\n",
"bugtrack_url": null,
"license": "Apache License\n Version 2.0, January 2004\n http://www.apache.org/licenses/\n \n TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION\n \n 1. Definitions.\n \n \"License\" shall mean the terms and conditions for use, reproduction,\n and distribution as defined by Sections 1 through 9 of this document.\n \n \"Licensor\" shall mean the copyright owner or entity authorized by\n the copyright owner that is granting the License.\n \n \"Legal Entity\" shall mean the union of the acting entity and all\n other entities that control, are controlled by, or are under common\n control with that entity. For the purposes of this definition,\n \"control\" means (i) the power, direct or indirect, to cause the\n direction or management of such entity, whether by contract or\n otherwise, or (ii) ownership of fifty percent (50%) or more of the\n outstanding shares, or (iii) beneficial ownership of such entity.\n \n \"You\" (or \"Your\") shall mean an individual or Legal Entity\n exercising permissions granted by this License.\n \n \"Source\" form shall mean the preferred form for making modifications,\n including but not limited to software source code, documentation\n source, and configuration files.\n \n \"Object\" form shall mean any form resulting from mechanical\n transformation or translation of a Source form, including but\n not limited to compiled object code, generated documentation,\n and conversions to other media types.\n \n \"Work\" shall mean the work of authorship, whether in Source or\n Object form, made available under the License, as indicated by a\n copyright notice that is included in or attached to the work\n (an example is provided in the Appendix below).\n \n \"Derivative Works\" shall mean any work, whether in Source or Object\n form, that is based on (or derived from) the Work and for which the\n editorial revisions, annotations, elaborations, or other modifications\n represent, as a whole, an original work of authorship. For the purposes\n of this License, Derivative Works shall not include works that remain\n separable from, or merely link (or bind by name) to the interfaces of,\n the Work and Derivative Works thereof.\n \n \"Contribution\" shall mean any work of authorship, including\n the original version of the Work and any modifications or additions\n to that Work or Derivative Works thereof, that is intentionally\n submitted to Licensor for inclusion in the Work by the copyright owner\n or by an individual or Legal Entity authorized to submit on behalf of\n the copyright owner. For the purposes of this definition, \"submitted\"\n means any form of electronic, verbal, or written communication sent\n to the Licensor or its representatives, including but not limited to\n communication on electronic mailing lists, source code control systems,\n and issue tracking systems that are managed by, or on behalf of, the\n Licensor for the purpose of discussing and improving the Work, but\n excluding communication that is conspicuously marked or otherwise\n designated in writing by the copyright owner as \"Not a Contribution.\"\n \n \"Contributor\" shall mean Licensor and any individual or Legal Entity\n on behalf of whom a Contribution has been received by Licensor and\n subsequently incorporated within the Work.\n \n 2. Grant of Copyright License. Subject to the terms and conditions of\n this License, each Contributor hereby grants to You a perpetual,\n worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n copyright license to reproduce, prepare Derivative Works of,\n publicly display, publicly perform, sublicense, and distribute the\n Work and such Derivative Works in Source or Object form.\n \n 3. Grant of Patent License. Subject to the terms and conditions of\n this License, each Contributor hereby grants to You a perpetual,\n worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n (except as stated in this section) patent license to make, have made,\n use, offer to sell, sell, import, and otherwise transfer the Work,\n where such license applies only to those patent claims licensable\n by such Contributor that are necessarily infringed by their\n Contribution(s) alone or by combination of their Contribution(s)\n with the Work to which such Contribution(s) was submitted. If You\n institute patent litigation against any entity (including a\n cross-claim or counterclaim in a lawsuit) alleging that the Work\n or a Contribution incorporated within the Work constitutes direct\n or contributory patent infringement, then any patent licenses\n granted to You under this License for that Work shall terminate\n as of the date such litigation is filed.\n \n 4. Redistribution. You may reproduce and distribute copies of the\n Work or Derivative Works thereof in any medium, with or without\n modifications, and in Source or Object form, provided that You\n meet the following conditions:\n \n (a) You must give any other recipients of the Work or\n Derivative Works a copy of this License; and\n \n (b) You must cause any modified files to carry prominent notices\n stating that You changed the files; and\n \n (c) You must retain, in the Source form of any Derivative Works\n that You distribute, all copyright, patent, trademark, and\n attribution notices from the Source form of the Work,\n excluding those notices that do not pertain to any part of\n the Derivative Works; and\n \n (d) If the Work includes a \"NOTICE\" text file as part of its\n distribution, then any Derivative Works that You distribute must\n include a readable copy of the attribution notices contained\n within such NOTICE file, excluding those notices that do not\n pertain to any part of the Derivative Works, in at least one\n of the following places: within a NOTICE text file distributed\n as part of the Derivative Works; within the Source form or\n documentation, if provided along with the Derivative Works; or,\n within a display generated by the Derivative Works, if and\n wherever such third-party notices normally appear. The contents\n of the NOTICE file are for informational purposes only and\n do not modify the License. You may add Your own attribution\n notices within Derivative Works that You distribute, alongside\n or as an addendum to the NOTICE text from the Work, provided\n that such additional attribution notices cannot be construed\n as modifying the License.\n \n You may add Your own copyright statement to Your modifications and\n may provide additional or different license terms and conditions\n for use, reproduction, or distribution of Your modifications, or\n for any such Derivative Works as a whole, provided Your use,\n reproduction, and distribution of the Work otherwise complies with\n the conditions stated in this License.\n \n 5. Submission of Contributions. Unless You explicitly state otherwise,\n any Contribution intentionally submitted for inclusion in the Work\n by You to the Licensor shall be under the terms and conditions of\n this License, without any additional terms or conditions.\n Notwithstanding the above, nothing herein shall supersede or modify\n the terms of any separate license agreement you may have executed\n with Licensor regarding such Contributions.\n \n 6. Trademarks. This License does not grant permission to use the trade\n names, trademarks, service marks, or product names of the Licensor,\n except as required for reasonable and customary use in describing the\n origin of the Work and reproducing the content of the NOTICE file.\n \n 7. Disclaimer of Warranty. Unless required by applicable law or\n agreed to in writing, Licensor provides the Work (and each\n Contributor provides its Contributions) on an \"AS IS\" BASIS,\n WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or\n implied, including, without limitation, any warranties or conditions\n of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A\n PARTICULAR PURPOSE. You are solely responsible for determining the\n appropriateness of using or redistributing the Work and assume any\n risks associated with Your exercise of permissions under this License.\n \n 8. Limitation of Liability. In no event and under no legal theory,\n whether in tort (including negligence), contract, or otherwise,\n unless required by applicable law (such as deliberate and grossly\n negligent acts) or agreed to in writing, shall any Contributor be\n liable to You for damages, including any direct, indirect, special,\n incidental, or consequential damages of any character arising as a\n result of this License or out of the use or inability to use the\n Work (including but not limited to damages for loss of goodwill,\n work stoppage, computer failure or malfunction, or any and all\n other commercial damages or losses), even if such Contributor\n has been advised of the possibility of such damages.\n \n 9. Accepting Warranty or Additional Liability. While redistributing\n the Work or Derivative Works thereof, You may choose to offer,\n and charge a fee for, acceptance of support, warranty, indemnity,\n or other liability obligations and/or rights consistent with this\n License. However, in accepting such obligations, You may act only\n on Your own behalf and on Your sole responsibility, not on behalf\n of any other Contributor, and only if You agree to indemnify,\n defend, and hold each Contributor harmless for any liability\n incurred by, or claims asserted against, such Contributor by reason\n of your accepting any such warranty or additional liability.\n \n END OF TERMS AND CONDITIONS\n \n APPENDIX: How to apply the Apache License to your work.\n \n To apply the Apache License to your work, attach the following\n boilerplate notice, with the fields enclosed by brackets \"[]\"\n replaced with your own identifying information. (Don't include\n the brackets!) The text should be enclosed in the appropriate\n comment syntax for the file format. We also recommend that a\n file or class name and description of purpose be included on the\n same \"printed page\" as the copyright notice for easier\n identification within third-party archives.\n \n Copyright [yyyy] [name of copyright owner]\n \n Licensed under the Apache License, Version 2.0 (the \"License\");\n you may not use this file except in compliance with the License.\n You may obtain a copy of the License at\n \n http://www.apache.org/licenses/LICENSE-2.0\n \n Unless required by applicable law or agreed to in writing, software\n distributed under the License is distributed on an \"AS IS\" BASIS,\n WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n See the License for the specific language governing permissions and\n limitations under the License.\n ",
"summary": "verl: Volcano Engine Reinforcement Learning for LLM",
"version": "0.4.1.dev0",
"project_urls": {
"Homepage": "https://github.com/girishbarca/verl-fork"
},
"split_keywords": [],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "7aa7288fd552b5cc2f60d4b911675f1d8649f6bcdeba4a36d538a58ba79df02c",
"md5": "ba996b27f1979d6315ec2cbd1d22d27d",
"sha256": "2578e3d338a26b2859c124089532f72db4f8905af69d7ce7c09ba55e83fa8d19"
},
"downloads": -1,
"filename": "verl_cgft-0.4.1.dev0-py3-none-any.whl",
"has_sig": false,
"md5_digest": "ba996b27f1979d6315ec2cbd1d22d27d",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.8",
"size": 709076,
"upload_time": "2025-07-27T17:30:44",
"upload_time_iso_8601": "2025-07-27T17:30:44.310440Z",
"url": "https://files.pythonhosted.org/packages/7a/a7/288fd552b5cc2f60d4b911675f1d8649f6bcdeba4a36d538a58ba79df02c/verl_cgft-0.4.1.dev0-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "267a6d38c181978c2d5026262cf35ccc0602f049be07fc19c97ae1829c1d0ecd",
"md5": "654136e2eec158637d159a0a591db79d",
"sha256": "7cf640233bc478e8200d9bf77dee3f90833626b8cdcd7ba8272ed9554ad9784e"
},
"downloads": -1,
"filename": "verl_cgft-0.4.1.dev0.tar.gz",
"has_sig": false,
"md5_digest": "654136e2eec158637d159a0a591db79d",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.8",
"size": 506629,
"upload_time": "2025-07-27T17:30:46",
"upload_time_iso_8601": "2025-07-27T17:30:46.304905Z",
"url": "https://files.pythonhosted.org/packages/26/7a/6d38c181978c2d5026262cf35ccc0602f049be07fc19c97ae1829c1d0ecd/verl_cgft-0.4.1.dev0.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-07-27 17:30:46",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "girishbarca",
"github_project": "verl-fork",
"travis_ci": false,
"coveralls": false,
"github_actions": true,
"requirements": [
{
"name": "accelerate",
"specs": []
},
{
"name": "codetiming",
"specs": []
},
{
"name": "datasets",
"specs": []
},
{
"name": "dill",
"specs": []
},
{
"name": "flash-attn",
"specs": []
},
{
"name": "hydra-core",
"specs": []
},
{
"name": "liger-kernel",
"specs": []
},
{
"name": "numpy",
"specs": [
[
"<",
"2.0.0"
]
]
},
{
"name": "pandas",
"specs": []
},
{
"name": "peft",
"specs": []
},
{
"name": "pyarrow",
"specs": [
[
">=",
"19.0.0"
]
]
},
{
"name": "pybind11",
"specs": []
},
{
"name": "pylatexenc",
"specs": []
},
{
"name": "pre-commit",
"specs": []
},
{
"name": "ray",
"specs": []
},
{
"name": "tensordict",
"specs": [
[
"<=",
"0.6.2"
]
]
},
{
"name": "torchdata",
"specs": []
},
{
"name": "transformers",
"specs": []
},
{
"name": "wandb",
"specs": []
},
{
"name": "packaging",
"specs": [
[
">=",
"20.0"
]
]
},
{
"name": "uvicorn",
"specs": []
},
{
"name": "fastapi",
"specs": []
},
{
"name": "latex2sympy2_extended",
"specs": []
},
{
"name": "math_verify",
"specs": []
}
],
"lcname": "verl-cgft"
}
