PyTorch_CIFAR10

Why we included this project

Anyone doing CIFAR-10 work who wants a pretrained classifier without training one from scratch will find this repo useful. It takes the official TorchVision CNN implementations, adjusts them for the 10-class CIFAR-10 task, and ships the resulting weights, so you can load a model and run inference right away. The README lists validation accuracy, parameter count, and file size for each network, which helps when you need to trade accuracy against memory. Training code runs on PyTorch-Lightning and is reproducible, so the same scripts work if you prefer to retrain or fine-tune. It is a handy baseline for people who want a known-good CIFAR-10 model to build on.