# facebookresearch/jepa

PyTorch code and models for V-JEPA self-supervised learning from video.

Repository: https://github.com/facebookresearch/jepa
Canonical: https://ross.abutalabs.com/products/jepa
Language: Python
License: NOASSERTION
License Family: other
Last push: 2025-02-27T23:15:27+00:00

## Health v2 (maintenance only)
Score: 29/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 8, release rhythm 35, longevity 66
- inputs: {"age_days": 933, "days_push": 552, "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 4105, forks 418 (observed 2026-08-28T04:08:35.630862+00:00)

## What it is
Official PyTorch implementation of V-JEPA, a self-supervised method for learning visual representations from video using a joint-embedding predictive architecture. It includes pretrained model weights and training/evaluation code for downstream video and image tasks.

## Use cases
- pretrain video encoders without labels
- extract video representations with a frozen backbone
- fine-tune or probe models on video classification tasks
- research self-supervised video learning methods
- decode latent video predictions to pixels with a diffusion model

## When to choose
- you need strong video representations without human annotations
- you want to reproduce or build on V-JEPA research
- you prefer frozen-backbone evaluation with lightweight probes

## When to avoid
- you need a production-ready video API rather than research code
- you lack GPU resources for large-scale pretraining
- you need text-supervised or multimodal models

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, computer-vision, video-processing
- domain: machine-learning, deep-learning, computer-vision
- platform: python
- tags: self-supervised-learning, video-representations, pytorch, pretrained-models, research-code, jepa, video, gpu

## Member repositories
- facebookresearch/jepa (main) score 29

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:35.630862+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 2026-08-29T18:23:14.563892+00:00, confidence not recorded.
  - readme: https://github.com/facebookresearch/jepa (fetched 2026-08-28T04:08:35.630862+00:00, sha f9f235fcaebb)
- Data as of 2026-08-30T08:39:29.467469+00:00.