# yitu-opensource/T2T-ViT

ICCV2021, Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet

Repository: https://github.com/yitu-opensource/T2T-ViT
Canonical: https://ross.abutalabs.com/products/t2t-vit
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Topics: vision-transformer, t2t-transformer, vit
Last push: 2023-10-27T19:00:24+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": 2048, "days_push": 1041, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1191, forks 177 (observed 2026-08-28T04:03:56.107728+00:00)

## What it is
Official PyTorch implementation of Tokens-to-Token ViT (T2T-ViT), an ICCV 2021 vision transformer architecture trained from scratch on ImageNet. It provides model definitions, pretrained checkpoints, and training scripts for the T2T-ViT family.

## Use cases
- train a vision transformer from scratch on ImageNet
- use pretrained T2T-ViT models for image classification
- compare vision transformer architectures by parameter count and accuracy
- fine-tune a T2T-ViT checkpoint on a custom image dataset
- reproduce ICCV 2021 T2T-ViT paper results

## When to choose
- you need a vision transformer that trains from scratch without large-scale pretraining
- you want pretrained ImageNet checkpoints with a good accuracy-to-parameter tradeoff
- you are researching progressive tokenization in transformers

## When to avoid
- you need a maintained general-purpose vision model zoo (use timm instead)
- you require support for the latest PyTorch versions or modern training features
- you need non-image modalities or production deployment tooling

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning
- domain: computer-vision, image-processing, deep-learning, machine-learning
- platform: python
- tags: vision-transformer, vit, image-classification, pytorch, imagenet, research-code, iccv-2021, gpu

## Member repositories
- yitu-opensource/T2T-ViT (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:56.107728+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 2026-08-30T06:22:59.727600+00:00, confidence not recorded.
  - readme: https://github.com/yitu-opensource/T2T-ViT (fetched 2026-08-28T04:03:56.107728+00:00, sha d40b257c23e6)
- Data as of 2026-08-30T08:39:29.467469+00:00.