# EdgeNeXt

**Category:** Efficient/Mobile · **Code:**
[`timm/models/edgenext.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/edgenext.py) ·
**Architectures:** 5 · **Pretrained weights:** 6 (6 with ImageNet-1k results)

EdgeNeXt is a hybrid CNN-Transformer for mobile devices that combines depthwise convolution encoders with split
depth-wise transpose attention (SDTA) encoders, which apply attention across channels.

## Papers

* [EdgeNeXt: Efficiently Amalgamated CNN-Transformer Architecture for Mobile Vision
  Applications](https://arxiv.org/abs/2206.10589)
* [Solving ImageNet: a Unified Scheme for Training any Backbone to Top Results](https://arxiv.org/abs/2204.03475)

## 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('edgenext_base.in21k_ft_in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `EdgeNeXt` |
|---|---|
| [`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 (`EdgeNeXt` with `edgenext_xx_small` at 256),
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 |
|---|---:|---:|---:|---:|
| `edgenext_base` | 18.51 | 3.85 | 256 | 2 |
| `edgenext_small` | 5.59 | 1.26 | 256 | 1 |
| `edgenext_small_rw` | 7.83 | 1.58 | 256 | 1 |
| `edgenext_x_small` | 2.34 | 0.54 | 256 | 1 |
| `edgenext_xx_small` | 1.33 | 0.26 | 256 | 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 |
|---|---:|---:|---:|---:|---|---|
| `edgenext_base.in21k_ft_in1k` | 256 (test 320) | 84.076 | 320 | 1000 | mit | [timm](https://huggingface.co/timm/edgenext_base.in21k_ft_in1k) |
| `edgenext_base.usi_in1k` | 256 (test 320) | 83.970 | 320 | 1000 | mit | [timm](https://huggingface.co/timm/edgenext_base.usi_in1k) |
| `edgenext_small.usi_in1k` | 256 (test 320) | 81.574 | 320 | 1000 | mit | [timm](https://huggingface.co/timm/edgenext_small.usi_in1k) |
| `edgenext_small_rw.sw_in1k` | 256 (test 320) | 80.452 | 320 | 1000 | mit | [timm](https://huggingface.co/timm/edgenext_small_rw.sw_in1k) |
| `edgenext_x_small.in1k` | 256 (test 288) | 75.706 | 288 | 1000 | mit | [timm](https://huggingface.co/timm/edgenext_x_small.in1k) |
| `edgenext_xx_small.in1k` | 256 (test 288) | 71.886 | 288 | 1000 | mit | [timm](https://huggingface.co/timm/edgenext_xx_small.in1k) |