# yuweihao/MambaOut

MambaOut: Do We Really Need Mamba for Vision? (CVPR 2025)

Repository: https://github.com/yuweihao/MambaOut
Canonical: https://ross.abutalabs.com/products/mambaout
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2025-03-09T11:15:26+00:00

## Health v2 (maintenance only)
Score: 19/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 10, release rhythm 8, longevity 60
- inputs: {"age_days": 842, "days_push": 542, "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 2704, forks 49 (observed 2026-08-28T04:07:11.540149+00:00)

## What it is
MambaOut is a PyTorch implementation of Gated CNN models from the CVPR 2025 paper 'MambaOut: Do We Really Need Mamba for Vision?', which questions whether state space models (Mamba) are necessary for vision tasks. It provides model definitions and pretrained ImageNet checkpoints showing Gated CNNs outperform visual Mamba models on image classification.

## Use cases
- train image classification models without mamba
- compare gated cnn vs vision mamba performance
- get pretrained imagenet checkpoints for gated cnn
- reproduce cvpr 2025 mambaout paper results
- use mambaout models in timm for transfer learning
- study whether ssm is needed for vision backbones

## When to choose
- you need a strong convolutional vision backbone without state space models
- you want to benchmark Mamba vs Gated CNN architectures on ImageNet
- you want lightweight pretrained models competitive with ViT-S

## When to avoid
- you specifically need sequence modeling or state space models for vision
- you need detection or segmentation backbones, which the paper does not cover
- you want a production-ready inference service rather than research code

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, image-processing
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: pytorch, image-classification, gated-cnn, state-space-models, research-code, cvpr-2025, pretrained-models, gpu

## Member repositories
- yuweihao/MambaOut (main) score 19

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:11.540149+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 2026-08-30T02:15:54.548160+00:00, confidence not recorded.
  - readme: https://github.com/yuweihao/MambaOut (fetched 2026-08-28T04:07:11.540149+00:00, sha ec4f11d7b9f2)
- Data as of 2026-08-30T08:39:29.467469+00:00.