• A self-contained colab notebook for playing with CSCG. No install required!





  • Most deep learning generative models use amortized inference. i.e, their encoders are trained to answer only a particular kind of query. How do we train graphical models to answer arbitrary queries?