Thank you very much for providing the paper and code regarding to PINO. I noticed in your paper that data and physical information can be sampled at different resolutions, and that physical information can be sampled at a higher resolution. However, in the train_2d file in the code, the train_2d_operator, data loss (myloss) and PDE loss (darcy_loss) both use the same pred, which may cause them to use the same grid precision. It may be necessary to redefine the pred used in darcy_loss.
Additionally, x and y are not directly given in the code, and I assume they need to be replaced with data_ic and u.
Thank you very much for providing the paper and code regarding to PINO. I noticed in your paper that data and physical information can be sampled at different resolutions, and that physical information can be sampled at a higher resolution. However, in the train_2d file in the code, the train_2d_operator, data loss (myloss) and PDE loss (darcy_loss) both use the same pred, which may cause them to use the same grid precision. It may be necessary to redefine the pred used in darcy_loss.
Additionally, x and y are not directly given in the code, and I assume they need to be replaced with data_ic and u.