Torch-RecHub: A Lightweight, Efficient, and Easy-to-use PyTorch Recommender Framework Torch-RecHub is a PyTorch-based framework for building recommender systems. It provides a modular design with support for 30+ recommendation models across ranking, matching, multi-task, and generative tasks. The framework includes standardized pipelines for data loading, training, and evaluation, with easy configuration via config files or command-line arguments. Key features include ONNX export for production deployment, cross-engine data processing (e.g., PySpark), and built-in experiment tracking with WandB, SwanLab, and TensorBoardX. It supports CPU, NVIDIA CUDA, AMD ROCm, and Huawei Ascend NPU hardware. Example scripts are provided for tasks like CTR prediction (e.g., DeepFM on Criteo), matching (e.g., DSSM on MovieLens), and multi-task learning (e.g., MMoE). The project structure separates core components (models, trainers, utilities) from examples, tutorials, and tests.