Microsoft Recommenders

Includes over 30 algorithms from classic ALS to state-of-the-art models like xDeepFM and LightGCN, as detailed in the algorithm table, covering collaborative filtering, content-based, and sequential approaches.

Notebooks guide through the full ML pipeline from data preparation to operationalization on Azure, providing practical best practices for deployment, as seen in examples like the SAR quick start.

Offers tools for evaluating algorithms with standard metrics like MAP and nDCG, demonstrated in the benchmark notebook that compares performance across multiple models.

Supports CPU, GPU, and PySpark environments through extras (e.g., [gpu], [spark]), allowing scaling based on computational needs, as mentioned in the setup guide.