This repository provides a complete PyTorch implementation of SimCLR (Simple Framework for Contrastive Learning of Visual Representations). It implements the contrastive learning approach where the model learns to bring similar images closer together in representation space while pushing dissimilar images apart. The implementation supports GPU training with mixed precision (AMP), linear evaluation protocol for feature quality assessment, and various datasets including STL10. It builds on torchvision for data augmentation transforms and ResNet backbones.