---
pipeline_tag: image-classification
license: apache-2.0
base_model: microsoft/beit-large-patch16-224
library_name: zeromodels
tags:
- keras
- zeromodels
- beit
- image-classification
- vit
- backbone
- arxiv:2106.08254
- pytorch
- jax
- tf
---

## ***See [our collection](https://huggingface.co/collections/zeromodels/beit-6a9352067192fd9fcfcfe6f1) for all versions of BEiT.***

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

[![GitHub](https://img.shields.io/badge/GitHub-ZeroModels-black?logo=github)](https://github.com/IMvision12/ZeroModels) [![Docs](https://img.shields.io/badge/Docs-BEiT-blue)](https://imvision12.github.io/ZeroModels/beit/) [![Collection](https://img.shields.io/badge/HF-BEiT%20collection-yellow)](https://huggingface.co/collections/zeromodels/beit-6a9352067192fd9fcfcfe6f1)

# zeromodels/beit-large-patch16-224

Paper: [BEiT: BERT Pre-Training of Image Transformers (arXiv:2106.08254)](https://arxiv.org/abs/2106.08254) · [HF Papers](https://huggingface.co/papers/2106.08254)

BEiT is a ViT-family vision transformer with a per-layer relative position bias, a learnable layer scale on each residual branch, and mean pooling of the patch tokens. Large backbone fine-tuned on ImageNet-1k (1000 classes).

For more details on the model, please go to Microsoft's original [model card](https://huggingface.co/microsoft/beit-large-patch16-224).

Pure-**Keras 3** conversion of [`microsoft/beit-large-patch16-224`](https://huggingface.co/microsoft/beit-large-patch16-224) for [zeromodels](https://github.com/IMvision12/ZeroModels). One implementation runs unmodified on **TensorFlow / Torch / JAX**.

This is a **image classification** checkpoint (`BeitImageClassify`).

## ✨ Quick start

```python
import os

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

from PIL import Image
from zeromodels.models.beit import BeitImageClassify, BeitModel, BeitImageProcessor

model = BeitImageClassify.from_weights("zeromodels/beit-large-patch16-224")
processor = BeitImageProcessor.from_weights("zeromodels/beit-large-patch16-224")

image = Image.open("your_image.jpg").convert("RGB")
pixels = processor(image)  # resize + normalize (normalization lives in the processor)
logits = model(pixels, training=False)
print(logits.shape)  # (1, num_classes)

# Feature extraction: the backbone without the classifier head
backbone = BeitModel.from_weights("zeromodels/beit-large-patch16-224", as_backbone=True)
features = backbone(pixels, training=False)
```

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

| Variant | Hub | Task |
|---|---|---|
| `beit-base-patch16-224` | [`zeromodels/beit-base-patch16-224`](https://huggingface.co/zeromodels/beit-base-patch16-224) | image classification |
| `beit-large-patch16-224` | [`zeromodels/beit-large-patch16-224`](https://huggingface.co/zeromodels/beit-large-patch16-224) | image classification |
| `beit-large-patch16-512` | [`zeromodels/beit-large-patch16-512`](https://huggingface.co/zeromodels/beit-large-patch16-512) | image classification |
| `beit-base-patch16-224-pt22k-ft22k` | [`zeromodels/beit-base-patch16-224-pt22k-ft22k`](https://huggingface.co/zeromodels/beit-base-patch16-224-pt22k-ft22k) | image classification |
| `beit-large-patch16-224-pt22k-ft22k` | [`zeromodels/beit-large-patch16-224-pt22k-ft22k`](https://huggingface.co/zeromodels/beit-large-patch16-224-pt22k-ft22k) | image classification |
| `beit-base-finetuned-ade-640-640` | [`zeromodels/beit-base-finetuned-ade-640-640`](https://huggingface.co/zeromodels/beit-base-finetuned-ade-640-640) | semantic segmentation |
| `beit-large-finetuned-ade-640-640` | [`zeromodels/beit-large-finetuned-ade-640-640`](https://huggingface.co/zeromodels/beit-large-finetuned-ade-640-640) | semantic segmentation |

## Tips

- Set `KERAS_BACKEND` **before** importing Keras / zeromodels.
- Normalization (0.5/0.5) is baked into the model, so pass raw `[0, 255]` pixels.
- Classification uses `BeitImageClassify`; semantic segmentation uses `BeitSemanticSegment` and returns logits at a quarter of the input resolution (upsample the `argmax` map to the input size).
- `BeitModel.from_weights(..., as_backbone=True)` returns the per-block token sequences for feature extraction.
- See [BEiT docs](https://imvision12.github.io/ZeroModels/beit/) and [Loading Weights](https://imvision12.github.io/ZeroModels/loading_weights/).
- Community / upstream safetensors still work via the `hf:` prefix, e.g. `BeitImageClassify.from_weights("hf:microsoft/beit-large-patch16-224")`.

## Special Thanks

A huge thank you to the Microsoft Research BEiT authors for creating and releasing these models.

License: Apache 2.0.