I am a software engineer at Emmi AI and we just released the open source framework we are using to train large-scale surrogate ML models.
Our aim was to lower the entry-barrier for Engineering AI, to get the power of modern deep learning into classical engineering.
Engineering and simulation data such as what is used for example in Computational Fluid Dynamics is of high-cardinality. Single examples can be up point clouds with multiple millions of points, requiring distinct approaches to models architectures and data loading that aren't served well by other frameworks. Noether comes with optimized architectures and components that make it easier to train models on simulation data.
Our aim was to lower the entry-barrier for Engineering AI, to get the power of modern deep learning into classical engineering.
Engineering and simulation data such as what is used for example in Computational Fluid Dynamics is of high-cardinality. Single examples can be up point clouds with multiple millions of points, requiring distinct approaches to models architectures and data loading that aren't served well by other frameworks. Noether comes with optimized architectures and components that make it easier to train models on simulation data.
We also provide a tutorial that runs you through all the steps involved in training a surrogate model: https://github.com/Emmi-AI/noether/blob/main/tutorial/README...