# SwiftFormer

**Category:** Efficient/Mobile · **Code:**
[`timm/models/swiftformer.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/swiftformer.py)
· **Architectures:** 4 · **Pretrained weights:** 4 (4 with ImageNet-1k results)

SwiftFormer replaces the quadratic matrix multiplications of self-attention with efficient additive attention, for
real-time mobile vision applications.

## Papers

* [SwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision
  Applications](https://arxiv.org/abs/2303.15446)

## 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('swiftformer_l3.dist_in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `SwiftFormer` |
|---|---|
| [`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 (`SwiftFormer` with `swiftformer_xs` 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 |
|---|---:|---:|---:|---:|
| `swiftformer_l1` | 12.06 | 1.60 | 224 | 1 |
| `swiftformer_l3` | 28.49 | 4.01 | 224 | 1 |
| `swiftformer_s` | 6.09 | 0.99 | 224 | 1 |
| `swiftformer_xs` | 3.48 | 0.61 | 224 | 1 |

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 |
|---|---:|---:|---:|---:|---|---|
| `swiftformer_l1.dist_in1k` | 224 | 80.906 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/swiftformer_l1.dist_in1k) |
| `swiftformer_l3.dist_in1k` | 224 | 82.994 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/swiftformer_l3.dist_in1k) |
| `swiftformer_s.dist_in1k` | 224 | 78.462 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/swiftformer_s.dist_in1k) |
| `swiftformer_xs.dist_in1k` | 224 | 75.600 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/swiftformer_xs.dist_in1k) |

## Citation

```bibtex
@InProceedings{Shaker_2023_ICCV,
    author    = {Shaker, Abdelrahman and Maaz, Muhammad and Rasheed, Hanoona and Khan, Salman and Yang, Ming-Hsuan and Khan, Fahad Shahbaz},
    title     = {SwiftFormer: Efficient Additive Attention for Transformer-based Real-time Mobile Vision Applications},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    year      = {2023},
}
```