InfoGAN | reproducing key results in the paper `` InfoGAN | Machine Learning library

python vqvae.py --train
                --n_embeddings [size of the latent space]
                --n_epochs     [number of epochs to train]
                --ema          [flag to use exponential moving average training for the embeddings]
           

# Retrain dSprites
dir=./quiz/public/models/dsprites
python src/train_dsprites.py --output_dir=$dir/ae  --variational=0
python src/train_dsprites.py --output_dir=$dir/vae --variational=1
python src/train_dsprites_supervised.py --output_dir=$dir/gt

#  

python draw.py -- train \
               -- use_read_attn \
               -- read_size=2 \
               -- use_write_attn \
               -- write_size=5 \
               -- [add'l options: e.g. n_epoch, z_size, lstm_size] \
               -- cud  

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# File: InfoGAN-mnist.py
# Author: Yuxin Wu

import argparse
import numpy as np
import os
import cv2
import tensorflow as tf

from tensorpack import *
from tensorpack.dataflow import dataset
from tensorpa