Callbacks and Loggers

API reference: pe.callback package and pe.logger package.

pe.callback.callback.Callback can be configured to be called after each Private Evolution iteration with the synthetic data as the input. It is useful for computing metrics, saving the synthetic samples, monitoring the progress, etc. Each pe.callback.callback.Callback can return a list of results (float numbers, images, matplotlib plots, etc.) in the form of pe.metric_item.MetricItem (see pe.metric_item). All pe.metric_item.MetricItem from all pe.callback.callback.Callback will be passed through each of the pe.logger.logger.Logger modules, which will then log the results in the desired way.

Available Callbacks

Currently, the following callbacks are implemented:

• For any data modality

  • pe.callback.common.compute_fid.ComputeFID: Computes the FID between the synthetic samples and the private samples.

  • pe.callback.common.save_checkpoints.SaveCheckpoints: Saves the checkpoint of current synthetic samples to files.

• Images

  • pe.callback.image.sample_images.SampleImages: Samples some images from each class.

  • pe.callback.image.save_all_images.SaveAllImages: Saves all synthetic images to files.

• Text

  • Coming soon!

Available Loggers

Currently, the following loggers are implemented:

• pe.logger.csv_print.CSVPrint: Saves the float results to a CSV file.

• pe.logger.log_print.LogPrint: Prints the float results to the console and/or files using the logging module.

• pe.logger.image_file.ImageFile: Saves the images to files.

• pe.logger.matplotlib_pdf.MatplotlibPDF: Saves the matplotlib plots to PDF files.