- cuda 11.8
# create new anaconda env
conda env create -f environment.yml
conda activate triton_v2
- The code structure is as follows:
./
|--networks
|--my_utils
|--exp
|--inference_simulation.sh
|--inference_simulation.py
|--inference_forecasting.sh
|--inference_forecasting.py
|--environment.yml
-
Download the whole
expfile from this link. -
Download the test data from this link1 and link2, and put them into a folder. (Note:
2020.h5in link1 andclimate_mean_s_t_ssh.npyare multi-part compressed files; you need to decompress them first.) -
Modify the following line in
my_utils.data_loader_multifiles.pyusing your own path.self.climate_mean = np.load('/apdcephfs_qy3/share_301734960/easyluwu/gy/data/coupled_0.25_23layers/climate_mean_s_t_ssh.npy', mmap_mode='r')
-
Modify the following line in
exp/Triton/20250819-012814/config.yamlusing your own path.exp_dir: '/apdcephfs_qy3/share_301734960/easyluwu/gy/triton_v2/exp_global_ocean_simulation_and_forecasting/exp'
test_data_path: '/apdcephfs_qy3/share_301734960/easyluwu/gy/data/coupled_0.25_23layers/test' (
2020.h5in this link1)test_data_path_atmos: '/jizhicfs/easyluwu/scaling_law/ft_local/weatherbench2/121_240/69var/test' (
2020.h5in this link2)land_mask_path: '/apdcephfs_qy3/share_301734960/easyluwu/gy/data/coupled_0.25_23layers/land_mask.h5'
global_means_path: '/apdcephfs_qy3/share_301734960/easyluwu/gy/data/coupled_0.25_23layers/mean_s_t_ssh.npy'
global_stds_path: '/apdcephfs_qy3/share_301734960/easyluwu/gy/data/coupled_0.25_23layers/std_s_t_ssh.npy'
global_means_path_atmos: '/jizhicfs/easyluwu/scaling_law/ft_local/weatherbench2/121_240/69var/mean.npy'
global_stds_path_atmos: '/jizhicfs/easyluwu/scaling_law/ft_local/weatherbench2/121_240/69var/std.npy'
Run the following script:
sh inference_simulation.sh
Run the following script:
sh inference_forecasting.sh