This repository adapts TorchVision’s popular CNN architectures for the CIFAR-10 dataset by modifying input dimensions, filter sizes, strides, and padding to accommodate the 32x32 images and 10 classes. It provides both the training code using PyTorch-Lightning and the trained model weights for immediate use. Supported models include VGG, ResNet, DenseNet, MobileNetV2, GoogleNet, and InceptionV3, with validation accuracies ranging from 92% to 94%.