Normalizing Flows

In this module are implemented so normalizing flows that can be used within the vae model or for sampling.

By convention, each implemented model is stored in a folder located in pythae.models.normalizing_flows and named likewise the model. The following modules can be found in this folder:

• modelname_config.py: Contains a ModelNameConfig instance inheriting from BaseNF.
• modelname_model.py: An implementation of the model inheriting either from BaseNFConfig

• modelname_utils.py (optional): A module where utils methods are stored.

All the codes are inspired from:

• https://github.com/kamenbliznashki/normalizing_flows
• https://github.com/karpathy/pytorch-normalizing-flows
• https://github.com/ikostrikov/pytorch-flows

Available Flows

BaseNF	Base Class from Normalizing flows
PlanarFlow	Planar Flow model.
RadialFlow	Radial Flow model.
MADE	Masked Autoencoder model
MAF	Masked Autoregressive Flow.
IAF	Inverse Autoregressive Flow.
PixelCNN	Pixel CNN model.