Lightning Talk: FTorch: A Library To Couple PyTorch ML Models With Fortran Climate Models - Matt Archer, University of Cambridge FTorch is open-source software that facilitates the coupling of PyTorch models with scientific codes written in Fortran. Many fields, such as climate science, fluid mechanics, and plasma physics, are increasingly using ML in research. PyTorch is Python-based, while many numerical models are written in Fortran, creating a challenge for researchers and engineers wanting to use ML. FTorch provides an intuitive, computationally efficient, and trusted approach for coupling PyTorch models into Fortran codes. It removes technical barriers so researchers can focus on applications and supports essential features like GPU acceleration. In this lightning talk, we’ll demonstrate how climate science projects can easily integrate PyTorch ML models with FTorch. We’ll highlight FTorch’s simplicity and exciting new developments like bringing autograd and online training to Fortran. Showcasing FTorch and its features will help connect the Fortran and PyTorch communities and broaden FTorch’s applications beyond climate science, fostering further development of the library.