lightgbm-org/LightGBM ? reverse-engineered prompt

I want to build a machine learning project using a really fast gradient boosting framework. It should use decision tree algorithms for tasks like classification, ranking, and regression. I've heard that LightGBM is great for this because it's super fast, doesn't use a ton of memory, and is often used in winning solutions for Kaggle competitions.

Please set up a project using this library. The core should be high-performance, but I'll mostly be using it from Python, so make sure there's a Python package ready to go. It should be able to handle large datasets and, if possible, be configured to use parallel processing or even my GPU to speed up the training process. The goal is to have a powerful tool for building accurate models quickly.