Oral Thu, May 21, 2026 • 3:15 PM – 3:30 PM PDT

AXLearn: Modular, Hardware-Agnostic Large Model Training

Mark Lee ⋅ Tom Gunter ⋅ Chang Lan ⋅ ⋅ Hanzhi Zhou ⋅ ⋅ Sneha Bangalore ⋅ ⋅ ⋅ Xianzhi Du ⋅ Philipp Dufter ⋅ ⋅ Ruixuan Hou ⋅ Haoshuo Huang ⋅ ⋅ Xiang Kong ⋅ Jinhao Lei ⋅ Tao Lei ⋅ Meng Li ⋅ Li Li ⋅ Jiarui Lu ⋅ Zhiyun Lu ⋅ ⋅ ⋅ ⋅ ⋅ Zhucheng Tu ⋅ Chong Wang ⋅ Jianyu Wang ⋅ ⋅ Zirui Wang ⋅ ⋅ Sam Wiseman ⋅ Guoli Yin ⋅ ⋅ Xiyou Zhou ⋅ Danyang Zhuo ⋅ ⋅ Ruoming Pang

Abstract

AXLearn is a production system which facilitates scalable and high-performance training of large deep learning models. Compared to other state-of-art deep learning systems, AXLearn has a unique focus on modularity and support for hardware-agnostic training. AXLearn's internal interfaces between software components follow strict encapsulation, allowing different components to be assembled to facilitate rapid model development and experimentation on different hardware infrastructure. AXLearn maintains constant complexity as we scale the components in the system, compared to linear or quadratic complexity in state-of-the-art training systems. This allows integrating features such as Rotary Position Embeddings (RoPE) into AXLearn across hundred of modules with just 10 lines of code, compared to hundreds as required in other systems. At the same time, AXLearn maintains equivalent performance compared to state-of-the-art training systems. Finally, we share our experience in the development and operation of AXLearn at Apple.