# allegro/allRank

allRank is a framework for training learning-to-rank neural models based on PyTorch.

Repository: https://github.com/allegro/allRank
Canonical: https://ross.abutalabs.com/products/allrank
Language: Python
License: Apache-2.0
License Family: permissive
Topics: learning-to-rank, ndcg, ranking, information-retrieval, pytorch, python, machine-learning, deep-learning, transformer, click-model
Last push: 2024-08-06T19:11:59+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2519, "days_push": 757, "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 1009, forks 131 (observed 2026-08-28T04:03:12.908004+00:00)

## What it is
allRank is a PyTorch-based framework for training neural learning-to-rank (LTR) models. It provides pointwise, pairwise, and listwise loss functions, Transformer and fully-connected scoring architectures, ranking metrics like NDCG and MRR, and click models for simulated click-through experiments.

## Use cases
- train a neural learning-to-rank model in pytorch
- experiment with listwise ranking losses like lambdaloss or approxndcg
- evaluate ranking models with ndcg and mrr metrics
- simulate click-through data with click models for LTR research
- train a transformer scoring function on libsvm ranking data

## When to choose
- you need a ready-made, configurable framework for neural LTR research or experimentation
- you want to compare many ranking loss functions under one training pipeline
- you work in PyTorch and want NDCG/MRR evaluation built in

## When to avoid
- you need a production search/ranking service rather than a training framework
- you want non-neural or gradient-boosted-tree LTR (e.g. XGBoost/LightGBM rankers)
- you need active feature development or frequent updates

## Facets
- artifact type: framework
- maturity: maintenance
- function: machine-learning, deep-learning, search-engine, benchmarking
- domain: machine-learning, deep-learning
- platform: python
- tags: learning-to-rank, pytorch, ndcg, ranking, loss-functions, transformer, click-models, libsvm, information-retrieval, search, docker, gpu, linux

## Member repositories
- allegro/allRank (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:12.908004+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 2026-08-30T07:12:01.298769+00:00, confidence not recorded.
  - readme: https://github.com/allegro/allRank (fetched 2026-08-28T04:03:12.908004+00:00, sha 81ab46394dff)
  - registry_pypi: https://pypi.org/pypi/allrank/json (fetched 2026-08-29T13:12:07.710778+00:00, sha e8d5d19bc00f)
- Data as of 2026-08-30T08:39:29.467469+00:00.