File size: 4,776 Bytes
2c3f465 d7d88c8 2c3f465 d7d88c8 2c3f465 d7d88c8 2c3f465 d7d88c8 2c3f465 d7d88c8 2c3f465 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 | ---
pipeline_tag: image-classification
license: apache-2.0
base_model: microsoft/beit-large-patch16-512
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
[](https://github.com/IMvision12/ZeroModels) [](https://imvision12.github.io/ZeroModels/beit/) [](https://huggingface.co/collections/zeromodels/beit-6a9352067192fd9fcfcfe6f1)
# zeromodels/beit-large-patch16-512
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 at 512x512 (1000 classes).
For more details on the model, please go to Microsoft's original [model card](https://huggingface.co/microsoft/beit-large-patch16-512).
Pure-**Keras 3** conversion of [`microsoft/beit-large-patch16-512`](https://huggingface.co/microsoft/beit-large-patch16-512) 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-patch16-512")
processor = BeitImageProcessor.from_weights("zeromodels/beit-large-patch16-512")
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-patch16-512", 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-patch16-512")`.
## Special Thanks
A huge thank you to the Microsoft Research BEiT authors for creating and releasing these models.
License: Apache 2.0.
|