lightfm

library1.17pypypiunverified

LightFM is a Python implementation of the LightFM hybrid recommendation algorithm, combining collaborative and content-based filtering. Current version is 1.17. Release cadence is irregular.

ai-ml

pip install lightfm

github ↗homepage ↗

INSTALL

IMPORT

SIG · LIGHTFM

lightfm

ai-mlpythonv1.17

harness data pending

Install & Compatibility

Where this runs

No compatibility data collected yet for this library.

Code

Verified usage

Verified import paths — ran on the pinned version, not inferred.

LightFM

✓ from lightfm import LightFM

cross_validate

✓ from lightfm.cross_validation import cross_validate

✗ from lightfm.evaluation import cross_validate

cross_validate moved to cross_validation submodule in v1.16

precision_at_k

✓ from lightfm.evaluation import precision_at_k

recall_at_k

✓ from lightfm.evaluation import recall_at_k

auc_score

✓ from lightfm.evaluation import auc_score

Basic usage: train a WARP-loss model on MovieLens data and evaluate precision@10.

import numpy as np from lightfm import LightFM from lightfm.datasets import fetch_movielens data = fetch_movielens(min_rating=4.0) model = LightFM(loss='warp') model.fit(data['train'], epochs=30, num_threads=2) from lightfm.evaluation import precision_at_k test_precision = precision_at_k(model, data['test'], k=10).mean() print(f'Test precision: {test_precision:.2f}')

Debug

Known issues

breakingIn v1.16, cross_validate was moved from lightfm.evaluation to lightfm.cross_validation. Old imports will break.

fix

Use 'from lightfm.cross_validation import cross_validate'.

affects: >=1.16

gotchaLightFM's fit method does not shuffle data internally. Always shuffle your training interactions (e.g., by using a random seed) to avoid order bias.

fix

Manually shuffle your item-user interaction matrix before passing to fit.

affects: all

gotchaThe predict method returns raw interaction scores (not probabilities). These can be negative and are not calibrated.

fix

Do not treat scores as probabilities; use ranking metrics for evaluation.

affects: all

deprecatedPython 3.6 and earlier are no longer supported. Ensure you use Python 3.7+.

fix

Upgrade to Python 3.7 or later.

affects: <=1.16

Upgrade

Version history

1.17latest on PyPI · released Mar 20, 2023

Audit

Dependencies

numpyrequiredrequired for array operations

scipyrequiredrequired for sparse matrix support

Agent activity

19 hits · last 30 days

node

Resources

githubhttps://github.com/lyst/lightfm ↗

homepagehttps://github.com/lyst/lightfm ↗