Data and code for training and testing a ResMLP model with experience replay for machine-learning physics parameterization

<p>This directory contains the training data and code for training and testing a ResMLP with experience replay for creating a machine-learning physics parameterization for the Community Atmospheric Model.&nbsp;</p> <p>The directory is structured as follows:</p> <p>1. Download training and testing data: https://portal.nersc.gov/archive/home/z/zhangtao/www/hybird_GCM_ML</p> <p>2. Unzip nncam_training.zip</p> <p>nncam_training</p> <p>&nbsp; &nbsp; - models</p> <p>&nbsp; &nbsp; &nbsp; &nbsp;model definition of ResMLP and other models for comparison purposes</p> <p>&nbsp; &nbsp; - dataloader&nbsp;</p> <p>&nbsp; &nbsp; &nbsp; &nbsp;utility scripts to load data into pytorch dataset</p> <p>&nbsp; &nbsp; - training_scripts</p> <p>&nbsp; &nbsp; &nbsp; &nbsp;scripts to train ResMLP model with/without experience replay</p> <p>&nbsp; &nbsp; - offline_test</p> <p>&nbsp; &nbsp; &nbsp; &nbsp;scripts to perform offline test (Table 2, Figure 2)</p> <p>3. Unzip nncam_coupling.zip</p> <p>nncam_srcmods</p> <p>&nbsp; &nbsp; &nbsp;- SourceMods</p> <p>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; SourceMods to be used with CAM modules for coupling with neural network</p> <p>&nbsp; &nbsp; &nbsp;- otherfiles</p> <p>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; additional configuration files to setup and run SPCAM with neural network</p> <p>&nbsp; &nbsp; &nbsp;- pythonfiles</p> <p>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp;python scripts to run neural network and couple with CAM</p> <p>&nbsp; &nbsp; &nbsp; - ClimAnalysis</p> <p>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; - paper_plots.ipynb</p> <p>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;scripts to produce online evaluation figures (Figure 1, Figure 3-10)</p> <p>&nbsp; &nbsp; &nbsp;&nbsp;</p>