# MViTv2

**Category:** ViT · **Code:**
[`timm/models/mvitv2.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/mvitv2.py) ·
**Architectures:** 8 · **Pretrained weights:** 7 (4 with ImageNet-1k results)

Multiscale Vision Transformers v2, a hierarchical transformer with pooling attention, decomposed relative position
embeddings and residual pooling connections.

## Papers

* [MViTv2: Improved Multiscale Vision Transformers for Classification and Detection](https://arxiv.org/abs/2112.01526)

## 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('mvitv2_tiny.fb_in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `MultiScaleVit` |
|---|---|
| [`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 (`MultiScaleVit` with `mvitv2_tiny` 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 |
|---|---:|---:|---:|---:|
| `mvitv2_base` | 51.47 | 10.16 | 224 | 1 |
| `mvitv2_base_cls` | 65.44 | 10.23 | 224 | 1 |
| `mvitv2_huge_cls` | 694.80 | 120.67 | 224 | 1 |
| `mvitv2_large` | 217.99 | 43.87 | 224 | 1 |
| `mvitv2_large_cls` | 234.58 | 42.17 | 224 | 1 |
| `mvitv2_small` | 34.87 | 7.00 | 224 | 1 |
| `mvitv2_small_cls` | 34.87 | 7.04 | 224 | 0 |
| `mvitv2_tiny` | 24.17 | 4.70 | 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 |
|---|---:|---:|---:|---:|---|---|
| `mvitv2_base.fb_in1k` | 224 | 84.424 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_base.fb_in1k) |
| `mvitv2_base_cls.fb_inw21k` | 224 | - | - | 19168 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_base_cls.fb_inw21k) |
| `mvitv2_huge_cls.fb_inw21k` | 224 | - | - | 19168 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_huge_cls.fb_inw21k) |
| `mvitv2_large.fb_in1k` | 224 | 85.264 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_large.fb_in1k) |
| `mvitv2_large_cls.fb_inw21k` | 224 | - | - | 19168 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_large_cls.fb_inw21k) |
| `mvitv2_small.fb_in1k` | 224 | 83.772 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_small.fb_in1k) |
| `mvitv2_tiny.fb_in1k` | 224 | 82.408 | 224 | 1000 | apache-2.0 | [timm](https://huggingface.co/timm/mvitv2_tiny.fb_in1k) |

## Citation

```bibtex
@inproceedings{li2021improved,
  title={MViTv2: Improved multiscale vision transformers for classification and detection},
  author={Li, Yanghao and Wu, Chao-Yuan and Fan, Haoqi and Mangalam, Karttikeya and Xiong, Bo and Malik, Jitendra and Feichtenhofer, Christoph},
  booktitle={CVPR},
  year={2022}
}
```