# MambaOut

**Category:** CNN · **Code:**
[`timm/models/mambaout.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/mambaout.py) ·
**Architectures:** 11 · **Pretrained weights:** 12 (11 with ImageNet-1k results)

MambaOut stacks Gated CNN blocks, Mamba blocks without the state space model (SSM) token mixer, to test whether SSMs are
needed for image classification.

## Papers

* [MambaOut: Do We Really Need Mamba for Vision?](https://arxiv.org/abs/2405.07992)

## Usage

See the [Quickstart](../quickstart) for loading pretrained weights, image preprocessing and fine-tuning, and [Feature
Extraction](../feature_extraction) for using the models as backbones.

```py
>>> import timm
>>> model = timm.create_model('mambaout_tiny.in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `MambaOut` |
|---|---|
| [`features_only=True` feature maps](../model_api#intermediate-features) | ✓ (strides 4, 8, 16, 32) |
| [`forward_intermediates()`](../model_api#intermediate-features) | ✓ |
| [`prune_intermediate_layers()`](../model_api#intermediate-features) | ✓ |
| [Any input size, no changes needed](../model_api#input-size) | ✓ |
| [`set_input_size()`](../model_api#input-size) | – |
| [`dynamic_img_size=True`](../model_api#input-size) | – |
| [`set_grad_checkpointing()`](../model_api#training-hooks) | ✓ |
| [Layer-wise LR decay (`group_matcher()`)](../model_api#training-hooks) | ✓ |
| [`reset_classifier()`](../model_api#classifier) | ✓ |

✓ supported, ✓ (square) works for square sizes only, ✗ not supported, – not needed (the model already accepts any input
size). Checked by running each feature on a randomly initialized model (`MambaOut` with `mambaout_femto` at 224), other
architectures of the same class are expected to behave the same. See the [Model API](../model_api) page for details.

## Architectures

| Architecture | Params (M) | GMACs | Input size | Pretrained weights |
|---|---:|---:|---:|---:|
| `mambaout_base` | 84.81 | 15.83 | 224 | 1 |
| `mambaout_base_plus_rw` | 101.66 | 19.19 | 224 | 3 |
| `mambaout_base_short_rw` | 88.83 | 16.31 | 224 | 1 |
| `mambaout_base_tall_rw` | 86.48 | 16.15 | 224 | 1 |
| `mambaout_base_wide_rw` | 94.45 | 17.78 | 224 | 1 |
| `mambaout_femto` | 7.30 | 1.16 | 224 | 1 |
| `mambaout_kobe` | 9.14 | 1.52 | 224 | 1 |
| `mambaout_small` | 48.49 | 8.96 | 224 | 1 |
| `mambaout_small_rw` | 48.50 | 8.96 | 224 | 1 |
| `mambaout_tiny` | 26.55 | 4.49 | 224 | 1 |
| `test_mambaout` | 0.45 | 0.03 | 160 | 0 |

GMACs and parameter counts are for the default input size and classifier of each architecture, from the [benchmark
results](https://github.com/huggingface/pytorch-image-models/tree/main/results).

## Pretrained weights

Top-1 is ImageNet-1k validation accuracy from
[`results-imagenet.csv`](https://github.com/huggingface/pytorch-image-models/blob/main/results/results-imagenet.csv) at
the listed eval size, - if the weight has no ImageNet-1k result (e.g. pretraining or image-text weights). Classes is the
size of the pretrained classifier, none for weights without one.

| Weight | Input size | Top-1 (%) | Eval size | Classes | License | Hub |
|---|---:|---:|---:|---:|---|---|
| `mambaout_base.in1k` | 224 (test 288) | 84.730 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base.in1k) |
| `mambaout_base_plus_rw.sw_e150_in12k` | 224 (test 288) | - | - | 11821 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base_plus_rw.sw_e150_in12k) |
| `mambaout_base_plus_rw.sw_e150_in12k_ft_in1k` | 224 (test 288) | 86.916 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base_plus_rw.sw_e150_in12k_ft_in1k) |
| `mambaout_base_plus_rw.sw_e150_r384_in12k_ft_in1k` | 384 | 87.508 | 384 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base_plus_rw.sw_e150_r384_in12k_ft_in1k) |
| `mambaout_base_short_rw.sw_e500_in1k` | 224 (test 288) | 84.842 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base_short_rw.sw_e500_in1k) |
| `mambaout_base_tall_rw.sw_e500_in1k` | 224 (test 288) | 84.988 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base_tall_rw.sw_e500_in1k) |
| `mambaout_base_wide_rw.sw_e500_in1k` | 224 (test 288) | 84.966 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_base_wide_rw.sw_e500_in1k) |
| `mambaout_femto.in1k` | 224 (test 288) | 79.912 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_femto.in1k) |
| `mambaout_kobe.in1k` | 224 (test 288) | 81.064 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_kobe.in1k) |
| `mambaout_small.in1k` | 224 (test 288) | 84.512 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_small.in1k) |
| `mambaout_small_rw.sw_e450_in1k` | 224 (test 288) | 84.604 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_small_rw.sw_e450_in1k) |
| `mambaout_tiny.in1k` | 224 (test 288) | 83.452 | 288 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mambaout_tiny.in1k) |