# facebookresearch/mmf

A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)

Repository: https://github.com/facebookresearch/mmf
Canonical: https://ross.abutalabs.com/products/mmf
Homepage: https://mmf.sh/
Language: Python
License: NOASSERTION
License Family: other
Topics: pytorch, vqa, pretrained-models, multimodal, deep-learning, captioning, dialog, textvqa, hateful-memes, multi-tasking
Last push: 2026-07-07T21:35:07+00:00

## Health v2 (maintenance only)
Score: 64/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 91, release rhythm 8, longevity 100
- inputs: {"age_days": 2989, "days_push": 57, "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 5633, forks 938 (observed 2026-08-28T04:09:26.462565+00:00)

## What it is
MMF is a modular PyTorch framework for vision and language multimodal research from Facebook AI Research. It ships reference implementations of state-of-the-art multimodal models, dataset tooling, and distributed training support, and serves as starter code for challenges like Hateful Memes, TextVQA, and VQA.

## Use cases
- train a visual question answering model
- run pretrained multimodal baselines like ViLT or M4C
- bootstrap a new vision-and-language research project
- fine-tune models on the Hateful Memes dataset
- benchmark on TextVQA, TextCaps, or VQA challenges
- add a custom multimodal dataset or model with minimal boilerplate

## When to choose
- you do vision-and-language multimodal research in PyTorch
- you want reference implementations of SOTA multimodal models
- you need distributed training and dataset/model zoos out of the box
- you are competing in VQA-family challenges

## When to avoid
- you need production multimodal inference serving rather than research code
- you work outside vision-and-language modalities
- you want a lightweight library without framework conventions
- you need actively maintained support for the latest model architectures

## Facets
- artifact type: framework
- maturity: maintenance
- function: machine-learning, deep-learning, nlp, image-processing, benchmarking
- domain: deep-learning, machine-learning, computer-vision
- platform: python, cross-platform
- tags: multimodal, vision-and-language, pytorch, vqa, captioning, pretrained-models, facebook-ai-research, natural-language-processing, research, gpu

## Member repositories
- facebookresearch/mmf (main) score 64

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:26.462565+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 2026-08-29T17:55:20.246093+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/mmf (fetched 2026-08-28T04:09:26.462565+00:00, sha 49b16676a36d)
  - homepage: https://mmf.sh/ (fetched 2026-08-29T08:49:58.659784+00:00, sha e217071df98d)
  - site_page: https://mmf.sh/docs (fetched 2026-08-29T08:49:58.668760+00:00, sha 0687f462fcbb)
- Data as of 2026-08-30T08:39:29.467469+00:00.