►
Keras 3 API documentation /
KerasNLP
KerasNLP
KerasNLP is a toolbox of modular building blocks ranging from pretrained state-of-the-art models, to low-level Transformer Encoder layers. For an introduction to the library see the KerasNLP home page. For a high-level introduction to the API see our getting started guide.
Pretrained Models
- Albert
- Bart
- Bert
- Bloom
- DebertaV3
- DistilBert
- Gemma
- Electra
- Falcon
- FNet
- GPT2
- Llama
- Llama3
- Mistral
- OPT
- PaliGemma
- Phi3
- Roberta
- XLMRoberta
Models API
- Backbone
- Task
- Preprocessor
- CausalLM
- CausalLMPreprocessor
- Seq2SeqLM
- Seq2SeqLMPreprocessor
- TextClassifier
- TextClassifierPreprocessor
- MaskedLM
- MaskedLMPreprocessor
- upload_preset
Tokenizers
- Tokenizer
- WordPieceTokenizer
- SentencePieceTokenizer
- BytePairTokenizer
- ByteTokenizer
- UnicodeCodepointTokenizer
- compute_word_piece_vocabulary function
- compute_sentence_piece_proto function
Preprocessing Layers
- StartEndPacker layer
- MultiSegmentPacker layer
- RandomSwap layer
- RandomDeletion layer
- MaskedLMMaskGenerator layer
Modeling Layers
- TransformerEncoder layer
- TransformerDecoder layer
- FNetEncoder layer
- PositionEmbedding layer
- RotaryEmbedding layer
- SinePositionEncoding layer
- ReversibleEmbedding layer
- TokenAndPositionEmbedding layer
- AlibiBias layer
- MaskedLMHead layer
- CachedMultiHeadAttention layer
Samplers
- Sampler base class
- BeamSampler
- ContrastiveSampler
- GreedySampler
- RandomSampler
- TopKSampler
- TopPSampler