# locuslab/convmixer

Implementation of ConvMixer for "Patches Are All You Need? 🤷"

Repository: https://github.com/locuslab/convmixer
Canonical: https://ross.abutalabs.com/products/convmixer
Language: Python
License: MIT
License Family: permissive
Last push: 2022-11-11T08:49:42+00:00

## Health v2 (maintenance only)
Score: 23/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 8, longevity 100
- inputs: {"age_days": 1800, "days_push": 1391, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1084, forks 98 (observed 2026-08-28T04:03:31.568594+00:00)

## What it is
An official PyTorch implementation of ConvMixer, a patch-based convolutional image classification architecture from the ICLR 2022 paper 'Patches Are All You Need?'. It includes model definitions, pretrained ImageNet weights, and training/evaluation scripts built on the timm framework.

## Use cases
- train a ConvMixer model on ImageNet
- evaluate pretrained ConvMixer checkpoints on ImageNet validation
- use ConvMixer as a backbone in timm
- reproduce results from the Patches Are All You Need paper
- train ConvMixer on CIFAR-10
- compare patch-based convnets against vision transformers

## When to choose
- you want a simple patch-based convolutional architecture for image classification
- you need pretrained ConvMixer weights for ImageNet experiments
- you use timm and want ConvMixer models integrated
- you are researching alternatives to vision transformers

## When to avoid
- you need a general-purpose production computer-vision pipeline
- you want actively developed state-of-the-art architectures
- you need non-PyTorch frameworks like TensorFlow or JAX

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, image-processing
- domain: deep-learning, computer-vision, image-processing
- platform: python
- tags: convmixer, pytorch, timm, image-classification, convolutional-networks, research-code, pretrained-weights, gpu

## Member repositories
- locuslab/convmixer (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:31.568594+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 2026-08-30T06:50:03.581088+00:00, confidence not recorded.
  - readme: https://github.com/locuslab/convmixer (fetched 2026-08-28T04:03:31.568594+00:00, sha 7af8d8f1fcb6)
- Data as of 2026-08-30T08:39:29.467469+00:00.