# datawhalechina/torch-rechub

A Lighting Pytorch Framework for Recommendation Models, Easy-to-use and Easy-to-extend.

Repository: https://github.com/datawhalechina/torch-rechub
Canonical: https://ross.abutalabs.com/products/torch-rechub
Homepage: https://datawhalechina.github.io/torch-rechub/
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: ctr-prediction, pytorch, recommendation-system, recsys, deep-learning, recommendation-algorithms, recommendation-engine, recommender-system, hstu, generative-recommendation, onnx, llm, ascend, npu
Last push: 2026-08-24T12:09:57+00:00

## Health v2 (maintenance only)
Score: 94/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 99, release rhythm 83, longevity 100
- inputs: {"age_days": 1574, "days_push": 9, "days_rel": 113, "gap_med": 17.0, "n_releases_24m": 11}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1207, forks 156 (observed 2026-08-28T04:03:59.518020+00:00)

## What it is
Torch-RecHub is a lightweight PyTorch framework for building recommendation system models with 30+ out-of-the-box algorithms covering ranking, matching, multi-task, and generative recommendation. It provides a standardized training/evaluation pipeline, ONNX export for deployment, and support for CPU, CUDA, ROCm, and Ascend NPU hardware.

## Use cases
- train ctr prediction models like deepfm and din in pytorch
- build a two-tower matching retrieval model
- run multi-task recommendation experiments like mmoe and ple
- export a trained recommendation model to onnx for serving
- benchmark recommendation algorithms on a unified pipeline
- learn recommendation systems with reproducible pytorch examples

## When to choose
- you want a simple, extensible pytorch framework for recommendation model experiments
- you need many classic and modern recsys models under one consistent trainer API
- you want easy ONNX export and hardware flexibility including Ascend NPU

## When to avoid
- you need a full production recommender serving platform rather than a modeling framework
- your stack is TensorFlow-based
- you need non-recommendation deep learning tasks

## Facets
- artifact type: framework
- maturity: active
- function: machine-learning, deep-learning, data-science, benchmarking
- domain: machine-learning, deep-learning, data-science
- platform: python, cross-platform
- tags: pytorch, recommender-system, ctr-prediction, deepfm, widedeep, din, multi-task-learning, onnx-export, matching, generative-recommendation, recsys, recommendation, recommendation-systems, gpu

## Member repositories
- datawhalechina/torch-rechub (main) score 94

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:59.518020+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 2026-08-30T06:18:58.249353+00:00, confidence not recorded.
  - readme: https://github.com/datawhalechina/torch-rechub (fetched 2026-08-28T04:03:59.518020+00:00, sha 09c12ae60cd2)
  - homepage: https://datawhalechina.github.io/torch-rechub/ (fetched 2026-08-29T12:26:45.944177+00:00, sha 3a7188900d9a)
  - registry_pypi: https://pypi.org/pypi/torch-rechub/json (fetched 2026-08-29T12:26:45.953303+00:00, sha 9a92cb7c88d4)
- Data as of 2026-08-30T08:39:29.467469+00:00.