Add option to provide a preprocess_bijection to train_flow. e.g. providing Invert(Affine(x.mean(axis=0), x.std(axis=0))) would perform standard scaling, and unscale the distribution by transforming the distribution with Invert(preprocess_bijection) after training.
Add option to provide a
preprocess_bijectiontotrain_flow. e.g. providingInvert(Affine(x.mean(axis=0), x.std(axis=0)))would perform standard scaling, and unscale the distribution by transforming the distribution withInvert(preprocess_bijection)after training.