# Gemma 4 ViT

**Category:** Encoder · **Code:**
[`timm/models/gemma4_vit.py`](https://github.com/huggingface/pytorch-image-models/blob/main/timm/models/gemma4_vit.py) ·
**Architectures:** 4 · **Pretrained weights:** 4 (0 with ImageNet-1k results)

The vision encoder of Google's Gemma 4 multimodal models, a ViT with 2D RoPE, gated MLPs, QK normalization and sandwich
normalization blocks.

## Papers

* [Gemma 4 Technical Report](https://arxiv.org/abs/2607.02770)

## 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('gemma4_vit_167m.gemma4_e4b_it', pretrained=True)
>>> data_config = timm.data.resolve_model_data_config(model)
```

## Model API support

| Feature | `Gemma4VitClassifier` | `Gemma4VitEncoder` |
|---|---|---|
| [`features_only=True` feature maps](../model_api#intermediate-features) | ✓ (strides 16, 16, 16) | ✓ (strides 16, 16, 16) |
| [`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 (`Gemma4VitClassifier` with `gemma4_vit_167m` at
256, `Gemma4VitEncoder` with `gemma4_vit_167m_enc` at 768), other architectures of the same class are expected to behave
the same. See the [Model API](../model_api) page for details.

## Architectures

| Architecture | Class | Params (M) | GMACs | Input size | Pretrained weights |
|---|---|---:|---:|---:|---:|
| `gemma4_vit_167m` | `Gemma4VitClassifier` | 167.36 | - | 768 | 1 |
| `gemma4_vit_167m_enc` | `Gemma4VitEncoder` | 167.36 | - | 768 | 1 |
| `gemma4_vit_570m` | `Gemma4VitClassifier` | 569.55 | - | 768 | 1 |
| `gemma4_vit_570m_enc` | `Gemma4VitEncoder` | 569.55 | - | 768 | 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 |
|---|---:|---:|---:|---:|---|---|
| `gemma4_vit_167m.gemma4_e4b_it` | 768 | - | - | none | apache-2.0 | [timm](https://huggingface.co/timm/gemma4_vit_167m.gemma4_e4b_it) |
| `gemma4_vit_167m_enc.gemma4_e4b_it` | 768 | - | - | none | apache-2.0 | [timm](https://huggingface.co/timm/gemma4_vit_167m_enc.gemma4_e4b_it) |
| `gemma4_vit_570m.gemma4_31b_it` | 768 | - | - | none | apache-2.0 | [timm](https://huggingface.co/timm/gemma4_vit_570m.gemma4_31b_it) |
| `gemma4_vit_570m_enc.gemma4_31b_it` | 768 | - | - | none | apache-2.0 | [timm](https://huggingface.co/timm/gemma4_vit_570m_enc.gemma4_31b_it) |