经 AI Skill Hub 精选评估，Relax 获评「强烈推荐」。这款Agent工作流在功能完整性、社区活跃度和易用性方面表现出色，AI 评分 8.0 分，适合有一定技术背景的用户使用。
Relax 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
Relax 是一套完整的 AI Agent 自动化工作流方案。通过可视化的节点编排，将复杂的多步骤任务拆解为清晰的自动化流程，实现全程无人值守的智能处理。支持与数百种外部服务和 API 无缝集成，适合构建数据处理管线、业务自动化和 AI 辅助决策系统。
# 方式一：pip 安装（推荐）
pip install relax
# 方式二：虚拟环境安装（推荐生产环境）
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install relax
# 方式三：从源码安装（获取最新功能）
git clone https://github.com/redai-infra/Relax
cd Relax
pip install -e .
# 验证安装
python -c "import relax; print('安装成功')"
# 命令行使用
relax --help
# 基本用法
relax input_file -o output_file
# Python 代码中调用
import relax
# 示例
result = relax.process("input")
print(result)
# relax 配置文件示例（config.yml） app: name: "relax" debug: false log_level: "INFO" # 运行时指定配置文件 relax --config config.yml # 或通过环境变量配置 export RELAX_API_KEY="your-key" export RELAX_OUTPUT_DIR="./output"
--balance-data with sub-batched forward to minimize GPU wasteray_native) and cross-cluster (external) federation modes______________________________________________________________________
The recommended way to run Relax is via the official Docker image, which ships with all CUDA, PyTorch, Megatron-LM, SGLang, and Ray dependencies pre-installed and version-matched.
```bash
Three end-to-end tasks cover text, vision-language, and omni-modal training. Each task downloads a public HuggingFace dataset and model, then launches training with a single script. Set EXP_DIR=/root (or wherever your models and datasets live) and the scripts will locate them automatically.
| Example | Description |
|---|---|
| [DeepEyes](./examples/deepeyes/) | Multi-modal vision-language RL with Qwen3-VL |
| [On-Policy Distillation](./examples/on_policy_distillation/) | Teacher-student knowledge distillation via KL penalty |
______________________________________________________________________
Relax is designed for omni-modal RL training — text, vision, and audio in one unified framework. Multimodal data is configured via the --multimodal-keys flag, with complete image/video/audio processing pipelines under relax/utils/multimodal/ for fine-grained control over image token counts, video frame sampling, and audio sample rates.
| Model Family | Sizes | Modality | Typical Tasks | Backend |
|---|---|---|---|---|
| **Qwen3** | 4B, 30B-A3B (MoE) | Text | Math reasoning, code, multi-turn dialogue, tool use | Megatron |
| **Qwen3-VL** | 4B, 30B-A3B | Vision + Language | Visual QA, image understanding, multimodal reasoning | Megatron |
| **Qwen3.5** | 30B-A3B | Vision + Language | Visual QA, image understanding, multimodal reasoning | Megatron |
| **Qwen3-Omni** | 30B-A3B | Text + Vision + Audio | Audio-visual QA, omni-modal understanding | Megatron |
| **Qwen3.6** | 35B-A3B (MoE) | Vision + Language | Visual QA, image understanding, multimodal reasoning | Megatron |
| **GLM5** | 744B-A40B (MoE) | Text | Math reasoning, code, multi-turn dialogue | Megatron |
| **Kimi K2.6** | ~1T-A32B (MoE) | Vision + Language | Visual QA, multimodal reasoning; INT4 QAT training | Megatron |
| **dots.mcore** | [dots.mocr](https://huggingface.co/rednote-hilab/dots.mocr) | Vision + Language | OCR, document understanding, multimodal reasoning | Megatron |
📖 New architectures are integrated via Megatron Bridge for automatic HF ↔ Megatron weight conversion.
______________________________________________________________________
Train Qwen3-Omni-30B-A3B on AVQA-R1-6K with GRPO and a multiple-choice reward.
hf download --repo-type dataset harryhsing/AVQA-R1-6K --local-dir /root/AVQA-R1-6K
hf download Qwen/Qwen3-Omni-30B-A3B-Instruct --local-dir /root/Qwen3-Omni-30B-A3B-Instruct
cd /root/Relax && export EXP_DIR=/root
bash -x scripts/entrypoint/spmd-multinode.sh \
scripts/training/multimodal/run-qwen3-30B-A3B-omni-16xgpu.shOnce running, you should see logs like:
Finish rollout 0/200
training step 0/200Checkpoints are saved in Megatron DCP format; convert them to HuggingFace weights with scripts/tools/convert_torch_dist_to_hf_bridge.py.
📖 Full walkthrough: Quick Start Guide · Customize Training · Configuration Guide
______________________________________________________________________
高性能的异步强化学习引擎
AI Skill Hub 为第三方内容聚合平台，本页面信息基于公开数据整理，不对工具功能和质量作任何法律背书。
建议在沙箱或测试环境中充分验证后，再部署至生产环境，并做好必要的安全评估。
✅ Apache 2.0 — 宽松开源协议，可商用，需保留版权声明和 NOTICE 文件，含专利授权条款。
AI Skill Hub 点评：Relax 的核心功能完整，质量优秀。对于自动化工程师和运维人员来说，这是一个值得纳入个人工具库的选择。建议先在非生产环境试用，再逐步推广。
| 原始名称 | Relax |
| Topics | aireinforcement-learningdistributed-training |
| GitHub | https://github.com/redai-infra/Relax |
| License | Apache-2.0 |
| 语言 | Python |
收录时间：2026-07-11 · 更新时间：2026-07-11 · License：Apache-2.0 · AI Skill Hub 不对第三方内容的准确性作法律背书。
选择 Agent 类型，复制安装指令后粘贴到对应客户端