How mhpi/hydrodl works
hydroDL is a hydrologic deep-learning toolkit for tasks such as streamflow forecasting and soil-moisture estimation. The main workflow is to load forcing, attribute, and observation data; prepare and normalize it; configure a model and loss; train; then validate or test predictions and analyze results. The repository also includes multiscale and uncertainty-oriented model/loss capabilities, CAMELS and CSV data access, and plotting/statistics utilities. Edges reflect the README and source-index references; several model and data modules were not sampled, so their internal connections are omitted.
Main entry points: dbCsv.py CSV data access · master.py Experiment orchestration · camels.py CAMELS access · train.py Model training
Components 17
Data access
- CAMELS access
hydroDL/data/camels.py - CSV data access
hydroDL/data/dbCsv.py - Data abstractions
hydroDL/dataset.py - Normalization
hydroDL/utils/norm.py
Modeling
- Recurrent models
hydroDL/model/rnn/CudnnLstmModel_R2P.py - CNN models
hydroDL/model/cnn/LstmCnn1d.py - Loss functions
hydroDL/model/crit.py - Dropout layers
hydroDL/model/dropout.py
Experiment workflow
- Experiment orchestration
hydroDL/master/master.py - Model training
hydroDL/model/train.py - Validation and testing
hydroDL/master/master.py - Multiscale training
hydroDL/model/train_multi.py
Results and utilities
- Scientific plotting
hydroDL/post/plot.py - Prediction statistics
hydroDL/post/stat.py - Time utilities
hydroDL/utils/time.py
Other components
- Hydrology researcher
- Hydrologic predictions
Connections 20
- Hydrology researcher → Experiment orchestration: configures experiment
README.md - Hydrology researcher → CAMELS access: selects data
README.md - Hydrology researcher → CSV data access: selects data
- CAMELS access → Data abstractions: uses abstractions
hydroDL/data/camels.py - CSV data access → Data abstractions: uses abstractions
hydroDL/data/dbCsv.py - CAMELS access → Time utilities: uses date ranges
hydroDL/data/camels.py - CSV data access → Time utilities: uses date ranges
hydroDL/data/dbCsv.py - CSV data access → Normalization: normalizes data
- Experiment orchestration → Model training: invokes training
hydroDL/master/master.py - Experiment orchestration → Validation and testing: invokes testing
hydroDL/master/master.py - Experiment orchestration → Recurrent models: configures model
hydroDL/master/master.py - Model training → Loss functions: uses loss
hydroDL/model/train.py - Model training → CNN models: uses CNN modules
hydroDL/model/train.py - CNN models → Dropout layers: uses dropout
hydroDL/model/cnn/LstmCnn1d.py - Recurrent models → Dropout layers: uses dropout
hydroDL/model/rnn/CudnnLstmModel_R2P.py - Validation and testing → Hydrologic predictions: produces predictions
hydroDL/master/master.py - Hydrology researcher → Multiscale training: runs multiscale workflow
README.md - Hydrology researcher → Scientific plotting: explores results
- Hydrology researcher → Prediction statistics: evaluates results
- Multiscale training → Prediction statistics: uses evaluation
hydroDL/model/train_multi.py