KerasHub Model Architectures

The following is a list of model architectures supported by KerasHub. These models can be created in two ways:

• Through the from_preset() constructor, which instantiates an object with a pre-trained configurations, vocabularies, and weights.
• Through custom configuration controlled by the user.

For the full list of available pretrained model presets shipped directly by the Keras team, see the Pretrained Models page.

Albert

• AlbertTokenizer
• AlbertBackbone model
• AlbertTextClassifier model
• AlbertTextClassifierPreprocessor layer
• AlbertMaskedLM model
• AlbertMaskedLMPreprocessor layer

Bart

• BartTokenizer
• BartBackbone model
• BartSeq2SeqLM model
• BartSeq2SeqLMPreprocessor layer

BASNet

• BASNetImageConverter
• BASNetBackbone model
• BASNetImageSegmenter model
• BASNetPreprocessor layer

Bert

• BertTokenizer
• BertBackbone model
• BertTextClassifier model
• BertTextClassifierPreprocessor layer
• BertMaskedLM model
• BertMaskedLMPreprocessor layer

Bloom

• BloomTokenizer
• BloomBackbone model
• BloomCausalLM model
• BloomCausalLMPreprocessor layer

CLIP

• CLIPTokenizer
• CLIPImageConverter
• CLIPBackbone model
• CLIPPreprocessor

CSPNet

• CSPNetImageConverter
• CSPNetBackbone model
• CSPNetImageClassifier model
• CSPNetImageClassifierPreprocessor layer

D-FINE

• DFineImageConverter
• DFineBackbone model
• DFineObjectDetector model
• DFineObjectDetectorPreprocessor layer

DebertaV3

• DebertaV3Tokenizer
• DebertaV3Backbone model
• DebertaV3TextClassifier model
• DebertaV3TextClassifierPreprocessor layer
• DebertaV3MaskedLM model
• DebertaV3MaskedLMPreprocessor layer

DeepLabV3

• DeepLabV3ImageConverter
• DeepLabV3Backbone model
• DeepLabV3ImageSegmenter model
• DeepLabV3ImageSegmenterPreprocessor layer

DeiT

• DeiTImageConverter
• DeiTBackbone model
• DeiTImageClassifier model
• DeiTImageClassifierPreprocessor layer

DenseNet

• DenseNetImageConverter
• DensNetBackbone model
• DenseNetImageClassifier model
• DenseNetImageClassifierPreprocessor layer

DepthAnything

• DepthAnythingImageConverter
• DepthAnythingBackbone model
• DepthAnythingDepthEstimator model

DINOV2

• DINOV2ImageConverter
• DINOV2Backbone model

DistilBert

• DistilBertTokenizer
• DistilBertBackbone model
• DistilBertTextClassifier model
• DistilBertTextClassifierPreprocessor layer
• DistilBertMaskedLM model
• DistilBertMaskedLMPreprocessor layer

EfficientNet

• EfficientNetImageConverter
• EfficientNetBackbone model
• EfficientNetImageClassifier model
• EfficientNetImageClassifierPreprocessor layer

Electra

• ElectraTokenizer
• ElectraBackbone model

ESM

• ESMTokenizer
• ESMBackbone model
• ESMProteinClassifier model
• ESMProteinClassifierPreprocessor layer
• ESMMaskedPLM model
• ESMMaskedPLMPreprocessor layer

Falcon

• FalconTokenizer
• FalconBackbone model
• FalconCausalLM model
• FalconCausalLMPreprocessor layer

Flux

• FluxBackbone model
• FluxTextToImage model
• FluxTextToImagePreprocessor layer

FNet

• FNetTokenizer
• FNetBackbone model
• FNetTextClassifier model
• FNetTextClassifierPreprocessor layer
• FNetMaskedLM model
• FNetMaskedLMPreprocessor layer

Gemma

• GemmaTokenizer
• GemmaBackbone model
• GemmaCausalLM model
• GemmaCausalLMPreprocessor layer

Gemma3

• Gemma3Tokenizer
• Gemma3ImageConverter
• Gemma3Backbone model
• Gemma3CausalLM model
• Gemma3CausalLMPreprocessor layer

GPT2

• GPT2Tokenizer
• GPT2Backbone model
• GPT2CausalLM model
• GPT2CausalLMPreprocessor layer

GPT-NeoX

• GPTNeoXTokenizer
• GPTNeoXBackbone model
• GPTNeoXCausalLM model
• GPTNeoXCausalLMPreprocessor layer

HGNetV2

• HGNetV2ImageConverter
• HGNetV2Backbone model
• HGNetV2ImageClassifier model
• HGNetV2ImageClassifierPreprocessor layer

Llama

• LlamaTokenizer
• LlamaBackbone model
• LlamaCausalLM model
• LlamaCausalLMPreprocessor layer

Llama3

• Llama3Tokenizer
• Llama3Backbone model
• Llama3CausalLM model
• Llama3CausalLMPreprocessor layer

Mistral

• MistralTokenizer
• MistralBackbone model
• MistralCausalLM model
• MistralCausalLMPreprocessor layer

MiT

• MiTImageConverter
• MiTBackbone model
• MiTImageClassifier model
• MiTImageClassifierPreprocessor layer

Mixtral

• MixtralTokenizer
• MixtralBackbone model
• MixtralCausalLM model
• MixtralCausalLMPreprocessor layer

MobileNet

• MobileNetImageConverter
• MobileNetBackbone model
• MobileNetImageClassifier model
• MobileNetImageClassifierPreprocessor layer

MobileNetV5

• MobileNetV5ImageConverter
• MobileNetV5Backbone model
• MobileNetV5ImageClassifier model
• MobileNetV5ImageClassifierPreprocessor layer

Moonshine

• MoonshineTokenizer
• MoonshineAudioConverter
• MoonshineBackbone model
• MoonshineAudioToText model
• MoonshineAudioToTextPreprocessor layer

OPT

• OPTTokenizer
• OPTBackbone model
• OPTCausalLM model
• OPTCausalLMPreprocessor layer

PaliGemma

• PaliGemmaTokenizer
• PaliGemmaBackbone model
• PaliGemmaCausalLM model
• PaliGemmaCausalLMPreprocessor layer

PARSeq

• PARSeqTokenizer
• PARSeqBackbone model
• PARSeqCausalLM model
• PARSeqCausalLMPreprocessor layer

Phi3

• Phi3Tokenizer
• Phi3Backbone model
• Phi3CausalLM model
• Phi3CausalLMPreprocessor layer

Qwen

• QwenTokenizer
• QwenBackbone model
• QwenCausalLM model
• QwenCausalLMPreprocessor layer

Qwen3

• Qwen3Tokenizer
• Qwen3Backbone model
• Qwen3CausalLM model
• Qwen3CausalLMPreprocessor layer

Qwen3Moe

• Qwen3MoeTokenizer
• Qwen3MoeBackbone model
• Qwen3MoeCausalLM model
• Qwen3MoeCausalLMPreprocessor layer

QwenMoe

• QwenMoeTokenizer
• QwenMoeBackbone model
• QwenMoeCausalLM model
• QwenMoeCausalLMPreprocessor layer

ResNet

• ResNetImageConverter
• ResNetBackbone model
• ResNetImageClassifier model
• ResNetImageClassifierPreprocessor layer

RetinaNet

• RetinaNetImageConverter
• RetinaNetBackbone model
• RetinaNetObjectDetector model
• RetinaNetObjectDetectorPreprocessor layer

Roberta

• RobertaTokenizer
• RobertaBackbone model
• RobertaTextClassifier model
• RobertaTextClassifierPreprocessor layer
• RobertaMaskedLM model
• RobertaMaskedLMPreprocessor layer

SegFormer

• SegFormerImageConverter
• SegFormerBackbone model
• SegFormerImageSegmenter model
• SegFormerImageSegmenterPreprocessor layer

Segment Anything Model

• SAMImageConverter
• SAMBackbone model
• SAMImageSegmenter model
• SAMImageSegmenterPreprocessor layer
• SAMPromptEncoder layer
• SAMMaskDecoder layer

SigLIP

• SigLIPTokenizer
• SigLIPImageConverter
• SigLIPBackbone model
• SigLIPPreprocessor

SmolLM3

• SmolLM3Tokenizer
• SmolLM3Backbone model
• SmolLM3CausalLM model
• SmolLM3CausalLMPreprocessor layer

Stable Diffusion 3

• SAMImageConverter
• StableDiffusion3Backbone model
• StableDiffusion3TextToImage model
• StableDiffusion3TextToImagePreprocessor layer
• StableDiffusion3ImageToImage model
• StableDiffusion3Inpaint model

T5

• T5Tokenizer
• T5Backbone model
• T5Preprocessor layer

T5Gemma

• T5GemmaTokenizer
• T5GemmaBackbone model
• T5GemmaSeq2SeqLM model
• T5GemmaSeq2SeqLMPreprocessor layer

VGG

• VGGImageConverter
• VGGBackbone model
• VGGImageClassifier model
• VGGImageClassifierPreprocessor layer

ViT

• ViTImageConverter
• ViTBackbone model
• ViTImageClassifier model
• ViTImageClassifierPreprocessor layer

ViTDet

• VitDet model

Whisper

• WhisperTokenizer
• WhisperAudioConverter
• WhisperBackbone model

Xception

• XceptionImageConverter
• XceptionBackbone model
• XceptionImageClassifier model
• XceptionImageClassifierPreprocessor layer

XLNet

• XLNetBackbone model

XLMRoberta

• XLMRobertaTokenizer
• XLMRobertaBackbone model
• XLMRobertaTextClassifier model
• XLMRobertaTextClassifierPreprocessor layer
• XLMRobertaMaskedLM model
• XLMRobertaMaskedLMPreprocessor layer