Matt Watson comments

Results 339 comments of


                                            Matt Watson

Add a vocab training utility for word piece

This is done!

Add a byte pair encoding (BPE) tokenizer layer

Closing this! We have an implementation released -> https://keras.io/api/keras_nlp/tokenizers/byte_pair_tokenizer/ If anyone encounters issue with the tokenizer, please file a bug!

Adding the ELECTRA Model

@shivance I actually think the thing to lead with here would be a "pre-training electra" example on keras.io. We could show the GAN-like setup used by ELECTRA with our low-level...

TransformerDecoder should support single inputs

This is fixed!

Integrate MLMMaskGenerator into BERT example

This is probably not a priority for us right now. Closing for now!

Add ragged support to SinePositionEncoding

Ah, this is indeed fixed! Thanks.

Add type annotations to the bert example

Keras as a whole is not pursuing type annotations at the moment.

Generate documentation for keras-nlp

This is done!

Adversarial Attacks and Testing the Robustness of Models

This probably does not quite fit with our current priorities.

Add checkpoints for fine-tuned models

@jbischof should we mark this as done?