The EvoX team has announced the official release of EvoX 1.0.0, a distributed GPU-accelerated evolutionary computation framework that now offers full compatibility with PyTorch. This major update transforms EvoX into a powerful, high-performance optimization tool for deep learning, reinforcement learning, and large-scale industrial applications.

What’s New in EvoX 1.0.0?

  • Full PyTorch Compatibility: EvoX now integrates seamlessly with the PyTorch ecosystem, making it easier than ever to apply evolutionary algorithms (EAs) in neural architecture search (NAS), reinforcement learning (RL), and meta-learning.

  • Distributed GPU Acceleration: Built for large-scale computation, EvoX leverages PyTorch for 100x speedup on heterogeneous hardware (CPUs, GPUs, multi-node clusters).

  • Extensive Algorithm Library: Features 50+ evolutionary algorithms, including GA, DE, PSO, CMA-ES, MOEAs (NSGA-II, RVEA, MOEA/D, etc.), and state-of-the-art meta-evolution methods.

  • RL & Physics Engine Support: Compatible with Brax and reinforcement learning environments, enabling evolutionary reinforcement learning (ERL) applications.

  • 100+ Benchmark Problems: Covers single-objective and multi-objective optimization, as well as real-world engineering challenges.

  • Customizable & Scalable: Supports flexible problem definitions, real-time data streaming, and scalable distributed workflows.

Bridging Evolutionary Computation and Deep Learning

EvoX 1.0.0 represents a groundbreaking step in merging evolutionary algorithms with modern deep learning frameworks. The integration with PyTorch enables researchers and practitioners to combine gradient-based learning with evolutionary search, unlocking new possibilities in AI-driven optimization, automated machine learning (AutoML), and complex decision-making systems.

Open-Source & Community-Driven

EvoX is now available on GitHub: https://github.com/EMI-Group/EvoX