The code of paper ‘High-Dimensional Latent Space Iterative Prediction: A Robust Framework for Enhanced Weather Forecasting Accuracy’.
To run the code, the following steps need to take:
-
The package need to install is in 'requirements.txt' file. The packages with the similar version are also OK.
-
ERA5 dataset at each time step should be stored as individual files and read by filename in the dataloader function (you need to download the ERA5 dataset in advance).
-
Adjust all 'paths' and run the code using the command:
training mode: `torchrun --nproc_per_node=2 --master_port 55562 main_latent.py --epoch 65 --ex_name 'baseline' --batch_size 16 --val_batch_size 32 --lr 1e-4 --test 0
testing mode: `torchrun --nproc_per_node=1 --master_port 55562 main_latent.py --epoch 65 --ex_name 'baseline' --batch_size 16 --val_batch_size 32 --lr 1e-4 --test 1