Manager

Manages initialization, preprocessing, training and anomaly detection of ecgan.

ecgan.manager.init()

Run initialization for ecgan and generate a config file.

ecgan.manager.preprocess()

Start preprocessing a dataset with parameters defined in the config file.

Requires a config file generated by ecgan-init. Raw data will be downloaded to the provided data path to '<data_path>/<dataset_name>/raw' in an arbitrary format given by the different datasets. The data will then be preprocessed and saved to '<data_path>/<dataset_name>/processed' in a format suitable for all other ecgan components.

ecgan.manager.train()

Start a training run.

Requires a config file generated by ecgan-init. Requires data preprocessed by ecgan-preprocess.

ecgan.manager.inverse()

Start an inverse mapping run for a trained model.

The inverse mapping requires an additional configuration. This configuration has to be generated using the init (-i) flag before the run can be started.

ecgan.manager.detect()

Start the anomaly detection process.

Requires generating an anomaly detection config file first using the -i`flag. `entity, project and run_name are mandatory during init to create the config file and not permitted as CLI parameters afterwards. Requires data preprocessed by ecgan-preprocess. Requires trained model from ecgan-train.