WideResNets for CIFAR10/100 implemented in PyTorch. This implementation requires less GPU memory than what is required by the official Torch implementation: https://github.com/szagoruyko/wide-residual-networks.
Example:
python train.py --dataset cifar100 --layers 40 --widen-factor 4
- densenet-pytorch
- Wide Residual Networks (BMVC 2016) http://arxiv.org/abs/1605.07146 by Sergey Zagoruyko and Nikos Komodakis.