How recommenders-team/recommenders works
Recommenders is a Python toolkit and example collection for prototyping, evaluating, tuning, and operationalizing recommendation systems. Its principal workflow is to prepare data, train a recommender, evaluate its predictions, and tune model parameters; operationalization is a documented production task. The package also supplies dataset loaders and splitters, a broad algorithm portfolio, and Python/Spark evaluation. The README establishes these product stages, while the excerpts show selected implementations; links to unsampled model families and operationalization are therefore documented capability links, not claims of inspected calls.
Main entry points: models Algorithm portfolio · datasets Dataset loaders · ncf_training.py NCF tuning run · python_evaluation.py Python metrics
Components 19
Data preparation
- Dataset loaders
recommenders/datasets - MovieLens loader
recommenders/datasets/movielens.py - Download helpers
recommenders/datasets/download_utils.py - DataFrame utilities
recommenders/datasets/pandas_df_utils.py - Data splitters
recommenders/datasets
Recommendation models
- Single-node SAR
recommenders/models/sar/sar_singlenode.py - Spark SAR+
contrib/sarplus/python/pysarplus/SARPlus.py - Neural CF
recommenders/models/ncf/ncf_singlenode.py - NCF dataset
recommenders/models/ncf/dataset.py - SASRec
recommenders/models/sasrec/model.py - Algorithm portfolio
recommenders/models
Evaluation and tuning
- Python metrics
recommenders/evaluation/python_evaluation.py - Spark metrics
recommenders/evaluation/spark_evaluation.py - Parameter tuning
recommenders/tuning - NCF tuning run
recommenders/tuning/nni/ncf_training.py - SVD tuning run
recommenders/tuning/nni/svd_training.py
Operations and utilities
- Operationalization
tools - Shared utilities
recommenders/utils
Other components
- Researcher / developer
Connections 19
- Researcher / developer → Dataset loaders: prepares data
README.md - Dataset loaders → MovieLens loader: loads MovieLens
README.md - MovieLens loader → Download helpers: downloads data
recommenders/datasets/movielens.py - Dataset loaders → Data splitters: splits data
README.md - Researcher / developer → Algorithm portfolio: builds model
README.md - Algorithm portfolio → Single-node SAR: includes SAR
README.md - Algorithm portfolio → Spark SAR+: includes Spark SAR
README.md - Algorithm portfolio → Neural CF: includes NCF
README.md - Algorithm portfolio → SASRec: includes sequential models
README.md - NCF tuning run → NCF dataset: constructs dataset
recommenders/tuning/nni/ncf_training.py - NCF tuning run → Neural CF: trains model
recommenders/tuning/nni/ncf_training.py - Researcher / developer → Python metrics: evaluates predictions
README.md - Researcher / developer → Spark metrics: evaluates Spark models
README.md - NCF tuning run → Python metrics: scores validation
recommenders/tuning/nni/ncf_training.py - SVD tuning run → Python metrics: scores validation
recommenders/tuning/nni/svd_training.py - Researcher / developer → Parameter tuning: tunes parameters
README.md - Researcher / developer → Operationalization: operationalizes model
README.md - Dataset loaders → DataFrame utilities: formats data
recommenders/datasets/pandas_df_utils.py - Single-node SAR → Shared utilities: uses constants
recommenders/models/sar/sar_singlenode.py