XBTorch: A Unified Framework for Modeling and Co-Design of Crossbar-Based Deep Learning Accelerators

Jan 11, 2026 by Osama Yousuf, Andreu L. Glasmann, Martin Lueker-Boden, Sina Najmaei, Gina C. Adam (arXiv.org)

DOI 10.48550/arXiv.2601.07086

Built XBTorch, a PyTorch-native toolkit that lets you plug realistic device models into crossbar DL accelerators and run hardware-aware training, co-design studies, and inference fault-tolerance experiments without reinventing the stack. It’s tech-agnostic (we show FeFETs) and gives you a unified, extensible simulation flow for device→system exploration — code on GitHub.

source S2, openalex