# pengzhiliang/MAE-pytorch

Unofficial PyTorch implementation of Masked Autoencoders Are Scalable Vision Learners

Repository: https://github.com/pengzhiliang/MAE-pytorch
Canonical: https://ross.abutalabs.com/products/mae-pytorch
Language: Python
License Family: other
Last push: 2023-07-25T06:31:32+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": 1754, "days_push": 1135, "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 2689, forks 341 (observed 2026-08-28T04:07:10.866090+00:00)

## What it is
An unofficial PyTorch implementation of the MAE (Masked Autoencoders Are Scalable Vision Learners) paper, providing pretraining and finetuning scripts for vision transformers. It implements masked token reconstruction with distributed training on ImageNet.

## Use cases
- pretrain a vision transformer with masked autoencoding
- reproduce MAE self-supervised pretraining in PyTorch
- finetune a masked autoencoder model on ImageNet
- learn how MAE masking and reconstruction works
- visualize MAE image reconstructions

## When to choose
- you want a PyTorch codebase for MAE-style self-supervised pretraining
- you want to study or modify the MAE masking and decoder implementation
- you need pretrain and finetune pipelines built on a BEiT-style codebase

## When to avoid
- you need guaranteed reproduction of the paper's reported results
- you want the official implementation with full feature parity (cls token, linear prob, knn eval are incomplete)
- you need a commercially licensed or maintained project (no license is provided)

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: masked-autoencoder, self-supervised-learning, vision-transformer, pytorch, unofficial-implementation, pretraining, gpu

## Member repositories
- pengzhiliang/MAE-pytorch (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:10.866090+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 2026-08-30T02:17:42.702816+00:00, confidence not recorded.
  - readme: https://github.com/pengzhiliang/MAE-pytorch (fetched 2026-08-28T04:07:10.866090+00:00, sha daa9439b0429)
- Data as of 2026-08-30T08:39:29.467469+00:00.