Here is a subset of the code for the paper.
To install requirements:
pip install -r requirements.txt
To use the notebook, you should also install the jupyter notebook kernels.
We provide training code for the following experiments:
- Inference over 2d latent space with Fisher divergence and KL divergence:
run_2d.py
,two_d.ipynb
- Training VAE model with objectives proposed in the paper:
run_ae.py
,denosing_vae.ipynb
- Bilevel optimization for VAE models:
run_bilevel_toy.py
All the scripts above should run without any arguments. See the argument list for specific settings.