keras-nlp

keras-nlp copied to clipboard

Make Decoding Functions Graph-compatible (with XLA Support!)

7

Resolves #241 - [x] Greedy Search - [ ] Beam Search (will probably open a separate PR for this) - [x] Top-p Search - [x] Top-k Search - [x] Random...

abheesht17

Allows easy cached download of the pre-trained vocabulary used to train BERT.

jessechancy

Support ragged inputs to our text generation utilites

2

For the text generation utilities listed [here](https://keras.io/api/keras_nlp/utils/), it would be nice to support ragged inputs, so we could remove the constraint that all seed text in a given batch needs...

mattdangerw

Implementation of Random Insertion Layer, ToDo: Tests, Docstring, Error Handling Creating the PR to discuss the API design and should we keep the insertion_list given it can be done through...

aflah02

Random Replacement

1

PR for Random Replacement Layer

aflah02

Fixes #223 @mattdangerw @chenmoneygithub I believe the PR is ready for review!

aflah02

SentencePieceTrainer

1

Fixes #249

aflah02

We need to add some workarounds for layers with a MultiHeadAttention layer to get build to function correctly. Fixes https://github.com/keras-team/keras-nlp/issues/270

mattdangerw

`char_to_token` in `keras_nlp.tokenizers.Tokenizer`

9

`char_to_token` is a method that converts the character index to the token index. See the HuggingFace method [here](https://huggingface.co/transformers/v3.2.0/main_classes/tokenizer.html#transformers.BatchEncoding.char_to_token) This is useful in span classification tasks, such as SQuaD, as we...

chenmoneygithub

Make an eval script for SQuaD

5

From a high level it is just a classification task, but there are some details to handle. The whole workflow can be described as: - Data 1. We can use...

chenmoneygithub