# RepGhost

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

RepGhostNet replaces the feature concatenation of the Ghost module with re-parameterized feature reuse, folded into a
single convolution path at inference.

## Papers

* [RepGhost: A Hardware-Efficient Ghost Module via Re-parameterization](https://arxiv.org/abs/2211.06088)

## 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('repghostnet_200.in1k', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `RepGhostNet` |
|---|---|
| [`features_only=True` feature maps](../model_api#intermediate-features) | ✓ (strides 2, 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 (`RepGhostNet` with `repghostnet_050` 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 |
|---|---:|---:|---:|---:|
| `repghostnet_050` | 2.31 | 0.05 | 224 | 1 |
| `repghostnet_058` | 2.54 | 0.06 | 224 | 1 |
| `repghostnet_080` | 3.27 | 0.10 | 224 | 1 |
| `repghostnet_100` | 4.06 | 0.15 | 224 | 1 |
| `repghostnet_111` | 4.52 | 0.18 | 224 | 1 |
| `repghostnet_130` | 5.46 | 0.24 | 224 | 1 |
| `repghostnet_150` | 6.55 | 0.31 | 224 | 1 |
| `repghostnet_200` | 9.77 | 0.53 | 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 |
|---|---:|---:|---:|---:|---|---|
| `repghostnet_050.in1k` | 224 | 66.978 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_050.in1k) |
| `repghostnet_058.in1k` | 224 | 68.938 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_058.in1k) |
| `repghostnet_080.in1k` | 224 | 72.240 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_080.in1k) |
| `repghostnet_100.in1k` | 224 | 74.208 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_100.in1k) |
| `repghostnet_111.in1k` | 224 | 75.060 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_111.in1k) |
| `repghostnet_130.in1k` | 224 | 76.378 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_130.in1k) |
| `repghostnet_150.in1k` | 224 | 77.474 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_150.in1k) |
| `repghostnet_200.in1k` | 224 | 78.802 | 224 | 1000 | mit | [timm](https://huggingface.co/timm/repghostnet_200.in1k) |