# meta-recsys/generative-recommenders

Repository hosting code for "Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations" (https://arxiv.org/abs/2402.17152).

Repository: https://github.com/meta-recsys/generative-recommenders
Canonical: https://ross.abutalabs.com/products/generative-recommenders
Language: Python
License: Apache-2.0
License Family: permissive
Topics: actions-speak-louder-than-words, recommendations, recsys, generative-recommenders, hstu, recommender-systems
Last push: 2026-08-24T21:32:07+00:00

## Health v2 (maintenance only)
Score: 69/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 35, longevity 62
- inputs: {"age_days": 877, "days_push": 9, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1966, forks 410 (observed 2026-08-28T04:05:59.881770+00:00)

## What it is
Meta's research library implementing HSTU and M-FALCON from the ICML'24 paper 'Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations'. It reformulates classical deep learning recommendation models (DLRMs) as generative modeling and provides efficient training/inference for large-scale sequential recommenders.

## Use cases
- train a generative sequential recommender model
- reproduce HSTU results on MovieLens and Amazon Reviews
- scale recommendation models to trillion parameters
- benchmark against SASRec, BERT4Rec, and GRU4Rec
- accelerate training and inference for large sequential recommenders
- study scaling laws for recommendation systems

## When to choose
- you need state-of-the-art generative sequential recommendation models
- you want to reproduce or extend the HSTU/M-FALCON paper results
- you are researching scaling laws for billion-user recommender systems

## When to avoid
- you need a production-ready plug-and-play recommender service
- you lack a GPU with 24GB+ memory
- you want a simple classical collaborative filtering solution

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-training, benchmarking, data-science
- domain: machine-learning, deep-learning, data-science
- platform: python
- tags: recommender-systems, recsys, hstu, generative-recommendations, sequential-modeling, transformers, pytorch, scaling-laws, research-code, icml-2024, algorithms, linux, gpu

## Member repositories
- meta-recsys/generative-recommenders (main) score 69

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:59.881770+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:05:27.325725+00:00, confidence not recorded.
  - readme: https://github.com/meta-recsys/generative-recommenders (fetched 2026-08-28T04:05:59.881770+00:00, sha 56d91c3c37bb)
- Data as of 2026-08-30T08:39:29.467469+00:00.