# dandelin/ViLT

Code for the ICML 2021 (long talk) paper: "ViLT: Vision-and-Language Transformer Without Convolution or Region Supervision"

Repository: https://github.com/dandelin/ViLT
Canonical: https://ross.abutalabs.com/products/vilt
Language: Python
License: Apache-2.0
License Family: permissive
Topics: vision-and-language
Last push: 2024-04-03T02:07:24+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": 1987, "days_push": 883, "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 1538, forks 229 (observed 2026-08-28T04:05:00.191120+00:00)

## What it is
Official PyTorch code for the ICML 2021 paper ViLT, a vision-and-language transformer that performs multimodal pre-training without convolutional image feature extractors or region supervision. It ships pretrained checkpoints and finetuned models for VQA, NLVR2, and image-text retrieval, plus Gradio demos.

## Use cases
- run visual question answering on images
- answer questions about an image with a pretrained model
- retrieve images from text descriptions
- finetune a vision-language transformer on my own dataset
- reproduce the ViLT paper results
- visualize masked language model attention over images

## When to choose
- you need a lightweight multimodal transformer without heavy object-detection feature extraction
- you want pretrained VQA or image-text retrieval models for research
- you want to reproduce or extend the ViLT paper

## When to avoid
- you need production-grade multimodal APIs with active support
- you want the latest state-of-the-art vision-language models
- you need a maintained library rather than research code

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, nlp, image-processing
- domain: artificial-intelligence, machine-learning, computer-vision
- platform: python
- tags: vision-and-language, transformer, vqa, visual-question-answering, image-text-retrieval, masked-language-modeling, pretrained-models, research-code, natural-language-processing, gpu

## Member repositories
- dandelin/ViLT (main) score 23

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:00.191120+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 2026-08-30T04:30:57.096307+00:00, confidence not recorded.
  - readme: https://github.com/dandelin/ViLT (fetched 2026-08-28T04:05:00.191120+00:00, sha e244820d32ac)
- Data as of 2026-08-30T08:39:29.467469+00:00.