DEGAS: Differentiable Efficient Generator Search

This project is for reproducing the results of paper DEGAS: Differentiable Efficient Generator Search

Prerequisites

Python 3.5.5,TensorFlow 1.4.0, NumPy, SciPy, Sklearn
To reproduce the IS and FID run: python train_gan.py dataset 'results/' labels --arch='arch_name' --gpu 1 --seed 1
- labels= 'unsup' or 'sup'
- For example, cifar10 evaluation:
  
  python train_gan.py cifar10 'results/' 'unsup' --arch='cifar10_n1_resnet_const_end_3e1_no_tg_200' --gpu 1 --seed 1

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
models		models
utils		utils
README.md		README.md
load_data.py		load_data.py
operations_found_models.py		operations_found_models.py
train_gan.py		train_gan.py