# idiap/fast-transformers

Pytorch library for fast transformer implementations

Repository: https://github.com/idiap/fast-transformers
Canonical: https://ross.abutalabs.com/products/fast-transformers
Language: Python
License Family: other
Last push: 2023-03-23T16:46:47+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2270, "days_push": 1259, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1779, forks 190 (observed 2026-08-28T04:05:35.264993+00:00)

## What it is
A PyTorch library providing fast transformer implementations with efficient attention mechanisms such as linear attention, designed to scale transformers to long sequences. It includes CUDA-accelerated kernels and builders for constructing transformer encoders with pluggable attention types.

## Use cases
- train transformers on long sequences without quadratic attention cost
- benchmark softmax attention vs linear attention speed
- implement research papers on fast attention mechanisms
- build transformer encoders with custom attention types in PyTorch
- speed up sequence encoding on GPU with CUDA kernels

## When to avoid
- you need a maintained library with active releases and support
- you want a general-purpose transformer toolkit like Hugging Face Transformers
- you cannot compile C++/CUDA extensions on your machine

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: machine-learning, deep-learning
- platform: python, windows
- tags: pytorch, transformers, attention, linear-attention, long-sequences, cuda, natural-language-processing, linux, macos, gpu

## Member repositories
- idiap/fast-transformers (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:35.264993+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 2026-08-30T03:24:39.225631+00:00, confidence not recorded.
  - readme: https://github.com/idiap/fast-transformers (fetched 2026-08-28T04:05:35.264993+00:00, sha 0d2f37c4cd01)
- Data as of 2026-08-30T08:39:29.467469+00:00.