This repository provides a Torch implementation of ResNet (Residual Networks), the deep convolutional neural network architecture from the 2015 paper by He et al. It includes training scripts for ResNet variants ranging from 18 to 200 layers on the ImageNet dataset, with support for multi-GPU training and NVIDIA cuDNN acceleration. Pre-trained model weights for all variants are available for download, and the code supports fine-tuning on custom datasets.