Hi, thank you for sharing the code for your paper!
I’m currently working to reproduce the results on the ERA5 dataset and have a few questions:
-
are the TU (total uncertainty) and L2 loss values in Figure 8 computed on the standardized data scale (e.g., temperature scaled to the [-1, 1] range)?
-
should I use the hyperparameters explicitly stated in the paper (e.g., resolution, batch size, epochs) or those defined in the repository’s makefile/configuration files?
Thank you for your time and guidance!
Hi, thank you for sharing the code for your paper!
I’m currently working to reproduce the results on the ERA5 dataset and have a few questions:
are the TU (total uncertainty) and L2 loss values in Figure 8 computed on the standardized data scale (e.g., temperature scaled to the [-1, 1] range)?
should I use the hyperparameters explicitly stated in the paper (e.g., resolution, batch size, epochs) or those defined in the repository’s makefile/configuration files?
Thank you for your time and guidance!