A framework for optimizing DSPy programs with RL.

---

🚀 Installation

Install Arbor via pip:

pip install -U arbor-ai

Optionally, you can also install:

pip install flash-attn --no-build-isolation

---

⚡ Quick Start

1️⃣ Make an arbor.yaml File

This is all dependent on your setup. Here is an example of one:

inference:
  gpu_ids: '0'

training:
  gpu_ids: '1, 2'

Which will use the GPU:0 for inference with GPU:1 and GPU:2 reserved for training. We generally recommend splitting the GPUs roughly evenly between inference and training.

2️⃣ Start the Server

CLI:

python -m arbor.cli serve --arbor-config arbor.yaml

3️⃣ Optimize a DSPy Program

Follow the DSPy tutorials here to see usage examples: DSPy RL Optimization Examples

---

🙏 Acknowledgements

Arbor builds on the shoulders of great work. We extend our thanks to:

• Will Brown's Verifiers library
• Hugging Face TRL library

📚 Citation

If you use this code in your research, please cite:

@misc{ziems2025arbor,
  title={Arbor: Open Source Language Model Post Training},
  author={Ziems, Noah and Agrawal, Lakshya A and Soylu, Dilara and Lai, Liheng and Miller, Isaac and Qian, Chen and Jiang, Meng and Khattab, Omar},
  year={2025}
}