CloudSat-based 3D Cloud Reflectivity Dataset and Transformer Model Code for Deep Learning Retrieval (2019–2020)

Description of files for the manuscript: "Adding depth to weather satellites: Reconstructing 3D cloud architecture from 2D views" - inputcld2019full.txt / inputcld2020full.txt: Input data files containing ABI radiance and auxiliary parameters for the years 2019 and 2020.- outputcld2019full.txt / outputcld2020full.txt: Corresponding output data files (ground truth) containing CloudSat radar reflectivity. - model.py: Contains the core Transformer model architecture.- loss.py: Defines the custom loss functions used for training.- class.py / reg.py: Main scripts for training the classification and regression models, respectively.- class_eval.py / reg_eval.py: Scripts for evaluating the trained models. Please refer to the Materials and Methods section of the manuscript for detailed descriptions.