Add a keras.io example using KerasNLP!

[mattdangerw]

KerasNLP is always looking for new examples on keras.io the demonstrate how to use the library.

This issue will stay open on a "contributions welcome" list forever. If you have an idea for a new example, feel free to leave a comment here to discuss, and then we can split out an issue to track the specific example once we have consensus.

[mattdangerw]

Development instructions for contributing examples are here.

Examples should be short <300 lines of code, and able to run fairly quickly with a GPU runtime in a colab. For many NLP problems, an example should show the idea behind a model architecture, without running full scale training.

[Plutone11011]

Hi @mattdangerw, do you think an example of Text Summarization might be interesting? It's also in the keras.io list of examples.

[Neeshamraghav012]

what about semantic search using KerasNLP @mattdangerw?

[jayam30]

Hey, I want to begin my contribution in Keras with this good first issue, please assign me

[shivance]

#1253 Semantic Similarity using KerasNLP

[atharvapurdue]

Hi @mattdangerw @jbischof @chenmoneygithub I had 2 examples which I feel might be a good example to contribute to keras.io

1. Example -

• BERT model for Chemical Reaction Yield Prediction (Regression Task)
• (reference - https://iopscience.iop.org/article/10.1088/2632-2153/abc81d ).
• The dataset which will be used is this - https://rxn4chemistry.github.io/rxn_yields/data/

I will make sure that the tutorial is easy to follow even for a non-domain expert

2. Example -

• Text Summerization using BART and Sampler functions (BART performs good on summerization)
• Dataset which will be used - https://www.kaggle.com/datasets/gowrishankarp/newspaper-text-summarization-cnn-dailymail

I am hoping that I can contribute more to this community. Please let me know if this is something that the community would like on keras.io Please let me know your thoughts

[taham655]

Hi, is there any good first issue that I can be assigned to?

[giuliocn]

Would you consider a new example on how to pretrain a backbone model with Keras Tuner ? Let me know what you think!

Example 1

• Data preprocessing for up to 1 Million examples.
• Model pre-training with Keras Tuner (e.g. RoBERTa)
• Model evaluation on a Text Classification task.
• (Optional) Comparison with a different pre-trained transformer.

[abuelnasr0]

Hi @mattdangerw @jbischof @chenmoneygithub. what about span labeling task for question answering by finetuning a RoBERTa model? there is a question-answering example in keras.io but it's not based on keras_nlp. what do you think?

[shivance]

Hi @mattdangerw @jbischof @chenmoneygithub. what about span labeling task for question answering by finetuning a RoBERTa model? there is a question-answering example in keras.io but it's not based on keras_nlp. what do you think?

Hey @abuelnasr0 , @TheAthleticCoder is working on it. See #741

[amangupta201]

@mattdangerw A Simple_MNIST_Covnet can also be a good idea https://github.com/amangupta201/keras-nlp/blob/master/Simple_MNIST_convnet.ipynb

[MrunaliG]

Hi, I would like to contribute an example that's a good first issue here. Would migrating the keras V2 example to V3 for Large-scale multi-label text classification https://keras.io/examples/nlp/multi_label_classification/ fall under this?