This repository contains the models for testing resnet networks on cifar-10
50000 samples for training, batch size 250, 200 iterations for 1 epoch, 64000 iterations in total
10000 samples for testing, test batch size 100 and 100 test iterations reduce learning rate after 32000 iters by factor of 10 then another factor of 10 after anohter 16000 iters
WRN-28-10 ["Wide Residual Networks" (http://arxiv.org/abs/1605.07146)]: best model achieved 0.958 accuracy on test datasets for single round and single crop