--- pipeline_tag: image-classification license: apache-2.0 base_model: timm/vit_small_patch16_224.augreg_in1k library_name: zeromodels tags: - keras - zeromodels - image-classification - vit - backbone - arxiv:2010.11929 - pytorch - jax - tf --- ## ***See [our collection](https://huggingface.co/collections/zeromodels/vit-6a8eae63ab67c7a19fb7a6da) for all versions of ViT.*** # Run ViT 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-ViT-blue)](https://imvision12.github.io/ZeroModels/classification_backbones/) [![Collection](https://img.shields.io/badge/HF-ViT%20collection-yellow)](https://huggingface.co/collections/zeromodels/vit-6a8eae63ab67c7a19fb7a6da) # zeromodels/vit_small_patch16_224_augreg_in1k Paper: [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale (arXiv:2010.11929)](https://arxiv.org/abs/2010.11929) · [HF Papers](https://huggingface.co/papers/2010.11929) Vision Transformer (ViT) patches an image and runs a transformer encoder. Use `ViTImageClassify` for logits or `ViTModel` for tokens / per-block features via `as_backbone=True`. For more details on the model, please go to the upstream [model card](https://huggingface.co/timm/vit_small_patch16_224.augreg_in1k). Pure-**Keras 3** conversion of [`timm/vit_small_patch16_224.augreg_in1k`](https://huggingface.co/timm/vit_small_patch16_224.augreg_in1k) for [zeromodels](https://github.com/IMvision12/ZeroModels). One implementation runs unmodified on **TensorFlow / Torch / JAX**. This is an **image-classification / backbone** checkpoint (`ViTImageClassify` / `ViTModel`). ## ✨ Quick start ```python import os os.environ["KERAS_BACKEND"] = "torch" # or "jax" / "tensorflow" from PIL import Image import numpy as np from zeromodels.models.vit import ViTImageClassify, ViTModel model = ViTImageClassify.from_weights("zeromodels/vit_small_patch16_224_augreg_in1k") backbone = ViTModel.from_weights( "zeromodels/vit_small_patch16_224_augreg_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 ViT variant the same way with `from_weights("zeromodels/")`: | Variant | Hub | |---|---| | `vit_base_patch16_224_augreg_in1k` | [`zeromodels/vit_base_patch16_224_augreg_in1k`](https://huggingface.co/zeromodels/vit_base_patch16_224_augreg_in1k) | | `vit_base_patch16_224_augreg_in21k` | [`zeromodels/vit_base_patch16_224_augreg_in21k`](https://huggingface.co/zeromodels/vit_base_patch16_224_augreg_in21k) | | `vit_base_patch16_224_augreg_in21k_ft_in1k` | [`zeromodels/vit_base_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_base_patch16_224_augreg_in21k_ft_in1k) | | `vit_base_patch16_224_orig_in21k_ft_in1k` | [`zeromodels/vit_base_patch16_224_orig_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_base_patch16_224_orig_in21k_ft_in1k) | | `vit_base_patch16_384_augreg_in1k` | [`zeromodels/vit_base_patch16_384_augreg_in1k`](https://huggingface.co/zeromodels/vit_base_patch16_384_augreg_in1k) | | `vit_base_patch16_384_augreg_in21k_ft_in1k` | [`zeromodels/vit_base_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_base_patch16_384_augreg_in21k_ft_in1k) | | `vit_base_patch16_384_orig_in21k_ft_in1k` | [`zeromodels/vit_base_patch16_384_orig_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_base_patch16_384_orig_in21k_ft_in1k) | | `vit_base_patch32_224_augreg_in1k` | [`zeromodels/vit_base_patch32_224_augreg_in1k`](https://huggingface.co/zeromodels/vit_base_patch32_224_augreg_in1k) | | `vit_base_patch32_224_augreg_in21k` | [`zeromodels/vit_base_patch32_224_augreg_in21k`](https://huggingface.co/zeromodels/vit_base_patch32_224_augreg_in21k) | | `vit_base_patch32_224_augreg_in21k_ft_in1k` | [`zeromodels/vit_base_patch32_224_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_base_patch32_224_augreg_in21k_ft_in1k) | | `vit_base_patch32_384_augreg_in1k` | [`zeromodels/vit_base_patch32_384_augreg_in1k`](https://huggingface.co/zeromodels/vit_base_patch32_384_augreg_in1k) | | `vit_base_patch32_384_augreg_in21k_ft_in1k` | [`zeromodels/vit_base_patch32_384_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_base_patch32_384_augreg_in21k_ft_in1k) | | `vit_large_patch16_224_augreg_in21k` | [`zeromodels/vit_large_patch16_224_augreg_in21k`](https://huggingface.co/zeromodels/vit_large_patch16_224_augreg_in21k) | | `vit_large_patch16_224_augreg_in21k_ft_in1k` | [`zeromodels/vit_large_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_large_patch16_224_augreg_in21k_ft_in1k) | | `vit_large_patch16_384_augreg_in21k_ft_in1k` | [`zeromodels/vit_large_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_large_patch16_384_augreg_in21k_ft_in1k) | | `vit_large_patch32_384_orig_in21k_ft_in1k` | [`zeromodels/vit_large_patch32_384_orig_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_large_patch32_384_orig_in21k_ft_in1k) | | `vit_small_patch16_224_augreg_in1k` | [`zeromodels/vit_small_patch16_224_augreg_in1k`](https://huggingface.co/zeromodels/vit_small_patch16_224_augreg_in1k) | | `vit_small_patch16_224_augreg_in21k` | [`zeromodels/vit_small_patch16_224_augreg_in21k`](https://huggingface.co/zeromodels/vit_small_patch16_224_augreg_in21k) | | `vit_small_patch16_224_augreg_in21k_ft_in1k` | [`zeromodels/vit_small_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_small_patch16_224_augreg_in21k_ft_in1k) | | `vit_small_patch16_384_augreg_in1k` | [`zeromodels/vit_small_patch16_384_augreg_in1k`](https://huggingface.co/zeromodels/vit_small_patch16_384_augreg_in1k) | | `vit_small_patch16_384_augreg_in21k_ft_in1k` | [`zeromodels/vit_small_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_small_patch16_384_augreg_in21k_ft_in1k) | | `vit_small_patch32_224_augreg_in21k` | [`zeromodels/vit_small_patch32_224_augreg_in21k`](https://huggingface.co/zeromodels/vit_small_patch32_224_augreg_in21k) | | `vit_small_patch32_224_augreg_in21k_ft_in1k` | [`zeromodels/vit_small_patch32_224_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_small_patch32_224_augreg_in21k_ft_in1k) | | `vit_small_patch32_384_augreg_in21k_ft_in1k` | [`zeromodels/vit_small_patch32_384_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_small_patch32_384_augreg_in21k_ft_in1k) | | `vit_tiny_patch16_224_augreg_in21k` | [`zeromodels/vit_tiny_patch16_224_augreg_in21k`](https://huggingface.co/zeromodels/vit_tiny_patch16_224_augreg_in21k) | | `vit_tiny_patch16_224_augreg_in21k_ft_in1k` | [`zeromodels/vit_tiny_patch16_224_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_tiny_patch16_224_augreg_in21k_ft_in1k) | | `vit_tiny_patch16_384_augreg_in21k_ft_in1k` | [`zeromodels/vit_tiny_patch16_384_augreg_in21k_ft_in1k`](https://huggingface.co/zeromodels/vit_tiny_patch16_384_augreg_in21k_ft_in1k) | ## Tips - Set `KERAS_BACKEND` **before** importing Keras / zeromodels. - `ViTImageClassify` returns class logits; `ViTModel` returns features (`as_backbone=True` for multi-scale stages). - See [docs](https://imvision12.github.io/ZeroModels/classification_backbones/) and [Loading Weights](https://imvision12.github.io/ZeroModels/loading_weights/). - Upstream / timm checkpoints: `ViTImageClassify.from_weights("hf:timm/vit_small_patch16_224.augreg_in1k")`. ## Special Thanks A huge thank you to the ViT authors and the timm / Hub communities for creating and releasing these models. License: see YAML `license` (usually matches the upstream checkpoint).