See our collection for all versions of DeiT.

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

kerasformers/deit3_huge_patch14_224_fb_in22k_ft_in1k

Paper: Training data-efficient image transformers and distillation through attention (arXiv:2012.12877) · HF Papers

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.

Pure-Keras 3 conversion of timm/deit3_huge_patch14_224.fb_in22k_ft_in1k for kerasformers. One implementation runs unmodified on TensorFlow / Torch / JAX.

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

✨ Quick start

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_huge_patch14_224_fb_in22k_ft_in1k")
backbone = DeiTModel.from_weights(
    "kerasformers/deit3_huge_patch14_224_fb_in22k_ft_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`
`deit3_base_patch16_224_fb_in22k_ft_in1k`	`kerasformers/deit3_base_patch16_224_fb_in22k_ft_in1k`
`deit3_base_patch16_384_fb_in1k`	`kerasformers/deit3_base_patch16_384_fb_in1k`
`deit3_base_patch16_384_fb_in22k_ft_in1k`	`kerasformers/deit3_base_patch16_384_fb_in22k_ft_in1k`
`deit3_huge_patch14_224_fb_in1k`	`kerasformers/deit3_huge_patch14_224_fb_in1k`
`deit3_huge_patch14_224_fb_in22k_ft_in1k`	`kerasformers/deit3_huge_patch14_224_fb_in22k_ft_in1k`
`deit3_large_patch16_224_fb_in1k`	`kerasformers/deit3_large_patch16_224_fb_in1k`
`deit3_large_patch16_224_fb_in22k_ft_in1k`	`kerasformers/deit3_large_patch16_224_fb_in22k_ft_in1k`
`deit3_large_patch16_384_fb_in1k`	`kerasformers/deit3_large_patch16_384_fb_in1k`
`deit3_large_patch16_384_fb_in22k_ft_in1k`	`kerasformers/deit3_large_patch16_384_fb_in22k_ft_in1k`
`deit3_medium_patch16_224_fb_in1k`	`kerasformers/deit3_medium_patch16_224_fb_in1k`
`deit3_medium_patch16_224_fb_in22k_ft_in1k`	`kerasformers/deit3_medium_patch16_224_fb_in22k_ft_in1k`
`deit3_small_patch16_224_fb_in1k`	`kerasformers/deit3_small_patch16_224_fb_in1k`
`deit3_small_patch16_224_fb_in22k_ft_in1k`	`kerasformers/deit3_small_patch16_224_fb_in22k_ft_in1k`
`deit3_small_patch16_384_fb_in1k`	`kerasformers/deit3_small_patch16_384_fb_in1k`
`deit3_small_patch16_384_fb_in22k_ft_in1k`	`kerasformers/deit3_small_patch16_384_fb_in22k_ft_in1k`
`deit_base_distilled_patch16_224_fb_in1k`	`kerasformers/deit_base_distilled_patch16_224_fb_in1k`
`deit_base_distilled_patch16_384_fb_in1k`	`kerasformers/deit_base_distilled_patch16_384_fb_in1k`
`deit_base_patch16_224_fb_in1k`	`kerasformers/deit_base_patch16_224_fb_in1k`
`deit_base_patch16_384_fb_in1k`	`kerasformers/deit_base_patch16_384_fb_in1k`
`deit_small_distilled_patch16_224_fb_in1k`	`kerasformers/deit_small_distilled_patch16_224_fb_in1k`
`deit_small_patch16_224_fb_in1k`	`kerasformers/deit_small_patch16_224_fb_in1k`
`deit_tiny_distilled_patch16_224_fb_in1k`	`kerasformers/deit_tiny_distilled_patch16_224_fb_in1k`
`deit_tiny_patch16_224_fb_in1k`	`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 and Loading Weights.
Upstream / timm checkpoints: DeiTImageClassify.from_weights("hf:timm/deit3_huge_patch14_224.fb_in22k_ft_in1k").