FTorch has been deployed in a number of scientific projects. This page collates the examples we are aware of. If you have a case study you'd like to see listed here, please get in touch or open a pull request to add it!
If you make use of FTorch in your work please cite our publication:
LS-DYNA explicit FEM solver through user-defined material subroutines -
DOI: 10.1016/j.cma.2026.119532GRIST-MPS - ML physics parameterisation suite trained on storm-resolving simulations,
coupled to the GRIST GCM for stable six-year AMIP runs -
DOI: 10.5194/gmd-19-5553-2026NLML neural network emulator of the exact nonlinear interactions, coupled online into
the WAVEWATCH III wave model -
DOI: 10.1029/2025JH000699ICON -
DOI: 10.1088/3049-4753/ae9e2eRemoTAP retrieval core -
DOI: 10.1109/TGRS.2026.3685881ICON) -
DOI: 10.1029/2025MS005544CAM,
imported from PyTorch via FTorch.
Trained on LES schemes and outperforming existing parameterisations -
DOI: 10.48550/arXiv.2511.01766 (preprint)E3SM.
See Hu et al. (2025) - DOI: 10.1029/2024MS004618 (and code)ICON giving a stable 20-year AMIP run -
DOI: 10.48550/arXiv.2510.08107 (preprint)CESM through learning model biases compared to ERA5\
DOI: 10.1029/2024GL114106WaveWatch III model -
DOI: 10.22541/essoar.174366388.80605654 (preprint)GloSea6 Seasonal Forecasting Model -
Replacing a BiCGStab bottleneck in the code with a deep learning approach to speed up execution without compromising model accuracy.
See Park and Chung (2025) - DOI: 10.3390/atmos16010060ICON atmospheric model
showing that best online performance occurs when causal relations are eliminated from the net.
See Heuer et al (2024) - DOI: 10.1029/2024MS004398CAM
atmospheric model.MiMA atmospheric model.
Demonstrates that nets trained near-identically offline can display greatly varied behaviours when coupled online.
See Mansfield and Sheshadri (2024) - DOI: 10.1029/2024MS004292