This part contains the setup for different experiments run to extend the baseline climatenet model.
Install and activate the environment
$ conda env create -f environment.yml
$ conda activate climatenet
The data is available at ClimateNet and must be downloaded manually. To retrain the model, the following folder hierarchy is required:
cl
|
└───data
|
└───train
|
└───val
|
└───test
All setup and hyperparameters are controlled via the climatenet_plus/config.json
file. Even though most variables have default values, the paths must be set in advance by each user. In order to do so, adjust the following entries in the configuration file:
Data directories
save_dir: <path to checkpoint folder>
data_dir: <path to data folder>
stats_dir: <path to feature_stats.json file >
Then run training via the following command
$ python main.py --model base mode train
or if on euler, run
$ sbatch run_slurm.sh
with the correct path to the data store
features
: choose from the 16 possible features in the ClimateNet Datasetarchitecture
: choose which model architecture to use. Possible models:cgnet
,unet
,upernet
,unetresnet
,segnet