# OFA-Sys/OFA

Official repository of OFA (ICML 2022). Paper: OFA: Unifying Architectures, Tasks, and Modalities Through a Simple Sequence-to-Sequence Learning Framework

Repository: https://github.com/OFA-Sys/OFA
Canonical: https://ross.abutalabs.com/products/ofa
Language: Python
License: Apache-2.0
License Family: permissive
Topics: multimodal, pretraining, image-captioning, text-to-image-synthesis, visual-question-answering, referring-expression-comprehension, vision-language, pretrained-models, prompt, prompt-tuning, chinese
Last push: 2024-04-24T06:20:34+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": 1677, "days_push": 861, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2557, forks 248 (observed 2026-08-28T04:07:00.823495+00:00)

## What it is
OFA is a unified sequence-to-sequence pretrained model supporting English and Chinese that unifies cross-modality, vision, and language tasks through a single framework. The repository provides training, finetuning, and prompt-tuning code plus checkpoints for tasks like image captioning, VQA, visual grounding, and text-to-image generation.

## Use cases
- generate image captions with a pretrained vision-language model
- run visual question answering on images
- perform visual grounding on referring expressions
- generate images from text descriptions
- finetune a unified multimodal seq2seq model
- do prompt tuning on a pretrained multimodal model
- classify images and text with one unified model

## When to choose
- you need a single pretrained model covering multiple vision-language tasks
- you want strong image captioning or VQA baselines with leaderboard-level checkpoints
- you need both English and Chinese multimodal support
- you want to experiment with prompt tuning on multimodal models

## When to avoid
- you need a lightweight production inference server rather than research code
- you want the latest large multimodal LLMs rather than a 2022-era seq2seq model
- you need tasks outside the supported vision-language set
- you require active development or frequent updates

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, nlp, image-processing, llm-training, prompt-engineering
- domain: artificial-intelligence, machine-learning, computer-vision, deep-learning
- platform: python
- tags: multimodal, vision-language, pretrained-models, sequence-to-sequence, image-captioning, visual-question-answering, text-to-image, prompt-tuning, research-code, natural-language-processing, gpu, linux

## Member repositories
- OFA-Sys/OFA (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:00.823495+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 2026-08-30T02:23:38.585931+00:00, confidence not recorded.
  - readme: https://github.com/OFA-Sys/OFA (fetched 2026-08-28T04:07:00.823495+00:00, sha a926f6197eda)
- Data as of 2026-08-30T08:39:29.467469+00:00.