keras-nlp

keras-nlp copied to clipboard

Add Flash Attention 2 support

5

**Is your feature request related to a problem? Please describe.** Flash Attention 2 is a library that provides attention operation kernels for faster and more memory efficient inference and training:...

rajveer43

Fix BytePair special tokens tokenization

2

BytePair already tokenize special tokens but it was having a small nit explained here keras-team/keras-nlp#1435 this PR fixes it.

abuelnasr0

high-level generation workflows (e.g. a LM class) on T5 (and Flan T5) is missing.

4

In Keras NLP **t5 model** the architecture and weights are present, but **HL workflows are missing** I would like to contribute a **high-level masked language modeling** workflow in the [t5...

debrupf2946

Cannot convert gpt2 generate into concrete function for tf lite

4

**Describe the bug** When I want to convert a generate function to a concrete function for tf lite, it errors out with `AttributeError: 'Tensor' object has no attribute 'numpy'`, the...

tianhaoz95

- Move the compiled while loop to the task base class. - Move as much common generative code to the task base classes. - Expose separate `prefill()` and `decode()` functions,...

mattdangerw

**Describe the bug** currently `BytePairTokenizer` supports tokenization of special tokens. but when it have two special tokens that have the same characters inside them like `` and ``, the tokenizer...

abuelnasr0

cannot import name 'CachedMultiHeadAttention' from partially initialized module 'keras_nlp.src.layers.modeling.cached_multi_head_attention' (most likely due to a circular import)

2

I am getting below error when i run my python script dealing with importing keras_nlp. **cannot import name 'CachedMultiHeadAttention' from partially initialized module 'keras_nlp.src.layers.modeling.cached_multi_head_attention' (most likely due to a circular...

anilmamidwar15021991

Potential RotaryEmbedding error?

3

Hi team, I was checking the implementation of `RotaryEmbedding` layer, and was a bit confused at the following computation: ``` def _apply_rotary_pos_emb(self, tensor, cos_emb, sin_emb): x1, x2 = ops.split(tensor, 2,...

chenmoneygithub

Some keras-nlp (keras 3) colabs stopped working

2

If I "Run all" [the Pretraining a Transformer from scratch with KerasNLP](https://keras.io/guides/keras_nlp/transformer_pretraining/) notebook in Colab, it fails at line: `tokenizer = keras_nlp.tokenizers.WordPieceTokenizer` with the following error: ```sh ImportError: WordPieceTokenizer requires...

rafaelsf80

Add a chat API to generative text models

1

Something similar to [this](https://ai.google.dev/tutorials/python_quickstart#chat_conversations) with a ChatSession that helps with the context. This feature was originally requested by Paul Mooney (Kaggle)

martin-gorner