A PyTorch Implementation of VaDE(https://arxiv.org/pdf/1611.05148.pdf)
First, execute pretrain.py to get parameters for initializing the model.
$ python pretrain.py --epochs 20 --gpu 0 --out parameters.pthThen, start training like
$ python main.py --epochs 300 --gpu 0 --pretrain parameters.pthThe results (accuracy and visualization of the latent space) can be confirmed by Tensorboard.
$ tensorboard --logdir runs