ICLR Reproducibility 2019: AutoLoss

papers/VelvetThunder-BJgK6iA5KX/README.txt

@@ -0,0 +1,46 @@
+=========================================
+Parth Kothari, Yuejiang Liu, Timur Lavrov
+Team: Velvet Thunder
+Project 2, Machine Learning, EPFL
+
+ICLR Reproducibility Challenge
+Issue
+https://github.com/reproducibility-challenge/iclr_2019/issues/89
+Codebase
+https://github.com/timur26/ICLR_Reproducibility_Challenge_Autoloss
+
+============= Dependency ================
+python 3.6
+tensorflow 1.11.0
+
+============= Download ================
+- Download MNIST data
+python download.py mnist
+- Download CIFAR10 data 
+Link: https://www.cs.toronto.edu/~kriz/cifar.html
+cd data 
+wget https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz 
+Unzip in 'cifar-10-python' folder as 'cifar10' folder 
+- Download MNIST Inception Score graph 
+mkdir model 
+cd model 
+wget https://github.com/tensorflow/models/raw/master/research/gan/mnist/data/classify_mnist_graph_def.pb
+
+============= Command ================
+- Joint minimization for classification:
+sh run_gs_classification.sh
+- Joint minimization for regression:
+sh run_gs_regression.sh
+- Hand-crafted schedule for classification:
+python c_main.py --lambdaa 1e-1 --mode "alter" --T 25
+- Auto-loss schedule for classification:
+python c_main.py --lambdaa 1e-1 --mode "autol" --T 25
+- GAN Baseline MNIST
+python main.py
+- GAN Baseline CIFAR-10
+python main.py --dataset cifar10
+- GAN Autoloss MNIST
+python main.py --autoloss
+- GAN Autoloss CIFAR10
+python main.py --autoloss --dataset cifar10
+