Transolver
A Transformer architecture for PDE surrogate modeling. It performs strongly on unstructured mesh solving tasks and is often used as a baseline comparison model.
Timeline (last 90 days)
The default model structure of this collision analysis recipe is a Transolver-type architecture based on graph neural networks.
Transolver is a geometry-oriented operator learning architecture that, building on the Transformer attention mechanism, introduces a physics-aware tokenization strategy that discretizes the input geometry into spatial slices and aggregates them into tokens.
Compared with the point-by-point message passing of traditional graph neural networks, Transolver can capture long-range dependencies on large-scale geometries more efficiently while remaining robust to mesh resolution.
Compared with Transolver on the same dataset and the same data split, the model reduces relative L2 error by 31% and interior mean squared error (interior MSE) by 47%.
Transolver is a transformer architecture that performs strongly in the field of PDE surrogate modeling.
All Facts (5)
The default model structure of this collision analysis recipe is a Transolver-type architecture based on graph neural networks.
50%UnverifiedTransolver is a geometry-oriented operator learning architecture that, building on the Transformer attention mechanism, introduces a physics-aware tokenization strategy that discretizes the input geometry into spatial slices and aggregates them into tokens.
50%UnverifiedCompared with the point-by-point message passing of traditional graph neural networks, Transolver can capture long-range dependencies on large-scale geometries more efficiently while remaining robust to mesh resolution.
50%UnverifiedCompared with Transolver on the same dataset and the same data split, the model reduces relative L2 error by 31% and interior mean squared error (interior MSE) by 47%.
50%UnverifiedTransolver is a transformer architecture that performs strongly in the field of PDE surrogate modeling.
50%