README.md

---
pipeline_tag: image-classification
license: apache-2.0
base_model: timm/deit3_base_patch16_384.fb_in1k
library_name: kerasformers
tags:
- keras
- kerasformers
- image-classification
- deit
- backbone
- arxiv:2012.12877
- pytorch
- jax
- tf
---

## ***See [our collection](https://huggingface.co/collections/kerasformers/deit-and-deit3-6a6bd40d28cbb9f69b422b5c) for all versions of DeiT.***

# Run DeiT with Keras 3: JAX, PyTorch, or TensorFlow

[![GitHub](https://img.shields.io/badge/GitHub-KerasFormers-black?logo=github)](https://github.com/IMvision12/KerasFormers) [![Docs](https://img.shields.io/badge/Docs-DeiT-blue)](https://imvision12.github.io/KerasFormers/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-DeiT%20collection-yellow)](https://huggingface.co/collections/kerasformers/deit-and-deit3-6a6bd40d28cbb9f69b422b5c)

# kerasformers/deit3_base_patch16_384_fb_in1k

Paper: [Training data-efficient image transformers and distillation through attention (arXiv:2012.12877)](https://arxiv.org/abs/2012.12877) · [HF Papers](https://huggingface.co/papers/2012.12877)

DeiT / DeiT3 are data-efficient ViT variants (distillation token in DeiT; improved training recipe in DeiT3). Same ImageClassify / Model API.

For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/deit3_base_patch16_384.fb_in1k).

Pure-**Keras 3** conversion of [`timm/deit3_base_patch16_384.fb_in1k`](https://huggingface.co/timm/deit3_base_patch16_384.fb_in1k) for [kerasformers](https://github.com/IMvision12/KerasFormers). One implementation runs unmodified on **TensorFlow / Torch / JAX**.

This is an **image-classification / backbone** checkpoint (`DeiTImageClassify` / `DeiTModel`).

## ✨ Quick start

```python
import os
os.environ["KERAS_BACKEND"] = "torch"  # or "jax" / "tensorflow"

from PIL import Image
import numpy as np
from kerasformers.models.deit import DeiTImageClassify, DeiTModel

model = DeiTImageClassify.from_weights("kerasformers/deit3_base_patch16_384_fb_in1k")
backbone = DeiTModel.from_weights(
    "kerasformers/deit3_base_patch16_384_fb_in1k", as_backbone=True
)

image = Image.open("your_image.jpg").convert("RGB")
image = image.resize((224, 224))
x = np.asarray(image, dtype="float32")[None]  # (1, H, W, 3)
print(model(x).shape)  # (1, num_classes)
feats = backbone(x)
print(len(feats), [tuple(f.shape) for f in feats])
```

Load any DeiT variant the same way with `from_weights("kerasformers/<variant>")`:

| Variant | Hub |
|---|---|
| `deit3_base_patch16_224_fb_in1k` | [`kerasformers/deit3_base_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit3_base_patch16_224_fb_in1k) |
| `deit3_base_patch16_224_fb_in22k_ft_in1k` | [`kerasformers/deit3_base_patch16_224_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_base_patch16_224_fb_in22k_ft_in1k) |
| `deit3_base_patch16_384_fb_in1k` | [`kerasformers/deit3_base_patch16_384_fb_in1k`](https://huggingface.co/kerasformers/deit3_base_patch16_384_fb_in1k) |
| `deit3_base_patch16_384_fb_in22k_ft_in1k` | [`kerasformers/deit3_base_patch16_384_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_base_patch16_384_fb_in22k_ft_in1k) |
| `deit3_huge_patch14_224_fb_in1k` | [`kerasformers/deit3_huge_patch14_224_fb_in1k`](https://huggingface.co/kerasformers/deit3_huge_patch14_224_fb_in1k) |
| `deit3_huge_patch14_224_fb_in22k_ft_in1k` | [`kerasformers/deit3_huge_patch14_224_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_huge_patch14_224_fb_in22k_ft_in1k) |
| `deit3_large_patch16_224_fb_in1k` | [`kerasformers/deit3_large_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit3_large_patch16_224_fb_in1k) |
| `deit3_large_patch16_224_fb_in22k_ft_in1k` | [`kerasformers/deit3_large_patch16_224_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_large_patch16_224_fb_in22k_ft_in1k) |
| `deit3_large_patch16_384_fb_in1k` | [`kerasformers/deit3_large_patch16_384_fb_in1k`](https://huggingface.co/kerasformers/deit3_large_patch16_384_fb_in1k) |
| `deit3_large_patch16_384_fb_in22k_ft_in1k` | [`kerasformers/deit3_large_patch16_384_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_large_patch16_384_fb_in22k_ft_in1k) |
| `deit3_medium_patch16_224_fb_in1k` | [`kerasformers/deit3_medium_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit3_medium_patch16_224_fb_in1k) |
| `deit3_medium_patch16_224_fb_in22k_ft_in1k` | [`kerasformers/deit3_medium_patch16_224_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_medium_patch16_224_fb_in22k_ft_in1k) |
| `deit3_small_patch16_224_fb_in1k` | [`kerasformers/deit3_small_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit3_small_patch16_224_fb_in1k) |
| `deit3_small_patch16_224_fb_in22k_ft_in1k` | [`kerasformers/deit3_small_patch16_224_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_small_patch16_224_fb_in22k_ft_in1k) |
| `deit3_small_patch16_384_fb_in1k` | [`kerasformers/deit3_small_patch16_384_fb_in1k`](https://huggingface.co/kerasformers/deit3_small_patch16_384_fb_in1k) |
| `deit3_small_patch16_384_fb_in22k_ft_in1k` | [`kerasformers/deit3_small_patch16_384_fb_in22k_ft_in1k`](https://huggingface.co/kerasformers/deit3_small_patch16_384_fb_in22k_ft_in1k) |
| `deit_base_distilled_patch16_224_fb_in1k` | [`kerasformers/deit_base_distilled_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit_base_distilled_patch16_224_fb_in1k) |
| `deit_base_distilled_patch16_384_fb_in1k` | [`kerasformers/deit_base_distilled_patch16_384_fb_in1k`](https://huggingface.co/kerasformers/deit_base_distilled_patch16_384_fb_in1k) |
| `deit_base_patch16_224_fb_in1k` | [`kerasformers/deit_base_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit_base_patch16_224_fb_in1k) |
| `deit_base_patch16_384_fb_in1k` | [`kerasformers/deit_base_patch16_384_fb_in1k`](https://huggingface.co/kerasformers/deit_base_patch16_384_fb_in1k) |
| `deit_small_distilled_patch16_224_fb_in1k` | [`kerasformers/deit_small_distilled_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit_small_distilled_patch16_224_fb_in1k) |
| `deit_small_patch16_224_fb_in1k` | [`kerasformers/deit_small_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit_small_patch16_224_fb_in1k) |
| `deit_tiny_distilled_patch16_224_fb_in1k` | [`kerasformers/deit_tiny_distilled_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit_tiny_distilled_patch16_224_fb_in1k) |
| `deit_tiny_patch16_224_fb_in1k` | [`kerasformers/deit_tiny_patch16_224_fb_in1k`](https://huggingface.co/kerasformers/deit_tiny_patch16_224_fb_in1k) |

## Tips

- Set `KERAS_BACKEND` **before** importing Keras / kerasformers.
- `DeiTImageClassify` returns class logits; `DeiTModel` returns features (`as_backbone=True` for multi-scale stages).
- See [docs](https://imvision12.github.io/KerasFormers/classification_backbones/) and [Loading Weights](https://imvision12.github.io/KerasFormers/loading_weights/).
- Upstream / timm checkpoints: `DeiTImageClassify.from_weights("hf:timm/deit3_base_patch16_384.fb_in1k")`.

## Special Thanks

A huge thank you to the DeiT authors and the timm / Hub communities for creating and releasing these models.

License: see YAML `license` (usually matches the upstream checkpoint).