Mixture of Experts using Discrete VAE
All files related to training the models reside in the code directory. To train the DMVAE model from scratch on MNIST dataset, simply run:
python train.py
This will automatically train the model and save the relevant reconstruction and generation plots. The parameters like model, dataset, etc can be controlled via command line arguments. To get a full list of all supported arguments, run:
python train.py --helpshort
Our code makes use of the following libraries:
- Tensorflow
- Numpy
- Sklearn
- Matplotlib
- tqdm