ScalingOpt - Optimization-Centric Efficient AI Community

<div align="center"> <img width="243.5" height="186.7" alt="2a0ff7d09549aec917655f98551eaa32" src="https://github.com/user-attachments/assets/d0a7d40e-1be4-4a34-a943-e19caef400d4" /> </div>

ScalingOpt is an optimization-centric community and knowledge hub for efficient AI. We focus on optimization methods, scalable algorithms, and resource-aware model design, with curated resources, tutorials, and research summaries.

Website: https://tianshijing.github.io/ScalingOpt/

If this repository is helpful, please consider giving it a star. Your support helps us reach more researchers and grow this resource.

Highlights

• A comprehensive optimizer directory with detailed metadata
• Research blogs and tutorials for modern optimization
• Benchmark summaries and practical training guidance
• Community-driven curation and continuous updates

News

[2026.3] Mano v2 is now officially released! Developed by Yufei Gu and Juanxi Tian, stay tuned for updates.

[2026.2] ScalingOpt's latest optimizer, Mano, is now officially released! Authored by Yufei Gu.

[2026.1] We are always welcoming new members to join us. ScalingOpt has now supplemented the A-Summary-Sheet-of-Optimization-in-Deep-Learning (collecting over 100 types of optimizers), authored by Yifeng Liu.

[2025.12] ScalingOpt team has published its first blog post: Backbone-Optimizer Coupling Bias: The Hidden Co-Design Principle. Welcome to read it!

[2025.11] ScalingOpt has been officially released and is currently undergoing continuous improvement and expansion of its content. Please stay tuned!

Community

Join our growing community of optimization researchers and practitioners.

• GitHub Discussions: Technical discussions and Q&A
• Research Collaboration: Connect with other researchers
• Blog Posts: Share your optimization insights
• Tutorial Contributions: Help others learn optimization

Acknowledgments

We thank the optimization research community for their groundbreaking work and contributions. Special thanks to:

• All researchers who developed the optimization algorithms featured in this platform

---