ZICQ
中 Log in / Sign up
Newsroom LLMs & Foundation Models #Hugging Face #AutoSella #Quantum Chemistry #AI Optimizer #Molecular Geometry Optimization

Hugging Face Releases AutoSella: AI-Driven Optimizer for Molecular Geometry Optimization

Avatar of Mr.Xu

By Mr.Xu Compiled & Reviewed by Editorial

Published:

中文阅读 (Chinese) English Version
Co-reading Select any sentence in the article to highlight, annotate, or copy a quote card. Join discussion ↓

Summary:Hugging Face has released AutoSella, an AI-driven optimizer designed to accelerate molecular geometry optimization in quantum chemistry workflows. By rewriting the fastest open-source optimizer, Sella, using a language model, AutoSella reduces the number of force calls without relying on DFT gradients. At the r2SCAN-3c DFT level, the best variant of AutoSella requires only 40.2% to 77.2% of Sella's force calls while achieving the same energy reduction. This innovation offers a more efficient sol


Background and Challenge

In quantum chemistry calculations, molecular geometry optimization is a critical and time-consuming step. Each optimization step requires one force evaluation, and at the density-functional theory (DFT) level, this evaluation dominates the computation time. While there are a variety of existing optimization methods, there is still room for improvement.

Technical Breakthrough

Hugging Face's research team has developed AutoSella, an AI-driven optimizer that rewrites existing optimizers using a language model to reduce the number of force calls. The development of AutoSella follows these key steps:

  1. Starting Point: Beginning with the fastest open-source optimizer available, Sella.
  2. Language Model Rewriting: Utilizing a language model to rewrite the optimizer to minimize force calls.
  3. Constraints: Implementing two admission gate mechanisms to reject premature stopping and improvements that do not generalize.
  4. Performance Validation: Validating the performance on molecular benchmarks and potential energy surfaces not used during the search to ensure generalization.

The experimental results show that AutoSella, at the r2SCAN-3c DFT level, significantly reduces the number of force calls, with the best variant requiring only 40.2% to 77.2% of Sella's force calls while achieving the same energy reduction.

Industry Impact

The release of AutoSella has the following impacts on the field of quantum chemistry:

  • Enhanced Computational Efficiency: By reducing the number of force calls, AutoSella significantly decreases computation time, particularly for complex molecular structures.
  • Resource Savings: Fewer force calls mean lower computational resource consumption, providing a more economical solution for large-scale quantum chemistry calculations.
  • Advancing AI in Scientific Computing: AutoSella demonstrates the potential of AI in optimizing complex scientific computations, paving the way for future AI-driven scientific research.

Developer Recommendations

For researchers and developers working in quantum chemistry, here are some recommendations:

  • Try AutoSella: Integrate AutoSella into existing quantum chemistry workflows to test its applicability to specific problems.
  • Stay Updated: Hugging Face may release more optimizer variants or improvements, so stay tuned for updates.
  • Explore AI in Scientific Computing: The success of AutoSella showcases the potential of AI in scientific computing, encouraging developers to explore more AI-driven scientific computation methods.

Source: Hugging Face Daily Papers (2026-10-05)

— END —

Tags: #Hugging Face #AutoSella #Quantum Chemistry #AI Optimizer #Molecular Geometry Optimization

Editorial & Fact-Checking Note: This article is compiled from primary research, official release documentation, and source papers by the ZICQ Newsroom pipeline with automated entity verification and human editorial review. If you notice any technical inaccuracy, please submit a correction via our corrections policy or email our editorial desk directly.

What's your take on this story?
🚀 Explore & Submit Works

Community Comments

Loading live comments and annotations…