# Twins

**Category:** ViT · **Code:**
[`timm/models/twins.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/twins.py) ·
**Architectures:** 6 · **Pretrained weights:** 6 (6 with ImageNet-1k results)

Twins-PCPVT (Pyramid Vision Transformer with conditional position encodings) and Twins-SVT (spatially separable
self-attention, locally grouped attention alternating with global sub-sampled attention).

## Papers

* [Twins: Revisiting the Design of Spatial Attention in Vision Transformers](https://arxiv.org/abs/2104.13840)

## 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('twins_pcpvt_small.in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `Twins` |
|---|---|
| [`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 (`Twins` with `twins_svt_small` 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 |
|---|---:|---:|---:|---:|
| `twins_pcpvt_base` | 43.83 | 6.68 | 224 | 1 |
| `twins_pcpvt_large` | 60.99 | 9.84 | 224 | 1 |
| `twins_pcpvt_small` | 24.11 | 3.83 | 224 | 1 |
| `twins_svt_base` | 56.07 | 8.59 | 224 | 1 |
| `twins_svt_large` | 99.27 | 15.15 | 224 | 1 |
| `twins_svt_small` | 24.06 | 2.94 | 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 |
|---|---:|---:|---:|---:|---|---|
| `twins_pcpvt_base.in1k` | 224 | 82.724 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/twins_pcpvt_base.in1k) |
| `twins_pcpvt_large.in1k` | 224 | 83.158 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/twins_pcpvt_large.in1k) |
| `twins_pcpvt_small.in1k` | 224 | 81.116 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/twins_pcpvt_small.in1k) |
| `twins_svt_base.in1k` | 224 | 83.122 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/twins_svt_base.in1k) |
| `twins_svt_large.in1k` | 224 | 83.718 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/twins_svt_large.in1k) |
| `twins_svt_small.in1k` | 224 | 81.694 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/twins_svt_small.in1k) |