geoprior.cli.stage1#

Stage-1: Zhongshan/Nansha preprocessing & sequence export for GeoPriorSubsNet

This script runs Steps 1–6 of the NATCOM pipeline:

  1. Load dataset

  2. Clean & select features

  3. Encode & scale

  4. Define feature sets

  5. Split by year & build PINN sequences

  6. Build train/val tf.data and EXPORT all arrays & metadata

Outputs#

  • CSVs: raw, cleaned, scaled

  • Joblib: one-hot encoder, main scaler, coord scaler

  • NPZ: train_inputs, train_targets, val_inputs, val_targets

  • JSON: manifest.json describing everything (paths, shapes, dims, columns, config), so Stage-2 can load without recomputing.

Stage-2 only needs to:

  • read manifest.json

  • np.load(…) the NPZs

  • joblib.load(…) the scalers/encoders if needed

  • build/compile/tune/train the model

Functions

main([argv])

run_stage1([overrides, config_root, config_path])

Execute Stage-1 preprocessing pipeline.

stage1_main([argv])

geoprior.cli.stage1.run_stage1(overrides=None, *, config_root='nat.com', config_path=None)[source]#

Execute Stage-1 preprocessing pipeline.

Parameters:
Return type:

None

geoprior.cli.stage1.stage1_main(argv=None)[source]#
Parameters:

argv (list[str] | None)

Return type:

None

geoprior.cli.stage1.main(argv=None)[source]#
Parameters:

argv (list[str] | None)

Return type:

None