keras-nlp VS Keras

Yes, Keras can be used to build LLMs. In fact this is one of the main use cases.

There are some tutorials about how to do it "from scratch", like this: https://keras.io/examples/nlp/neural_machine_translation_wit...

Otherwise, if you want to reuse an existing LLM (or just see how a large one would be implemented in practice) you can check out the models from KerasNLP. For instance, this is BERT, basically just a stack of TransformerEncoders. https://github.com/keras-team/keras-nlp/blob/master/keras_nl...

Yes, you can check out KerasCV and KerasNLP which host pretrained models like ResNet, BERT, and many more. They run on all backends as of the latest releases (today), and converting them to be backend-agnostic was pretty smooth! It took a couple of weeks to convert the whole packages.

https://github.com/keras-team/keras-nlp/tree/master/keras_nl...

Keras

As a beginner, I was looking for something simple and flexible for developing deep learning models and that is when I found Keras. Many AI/ML professionals appreciate Keras for its simplicity and efficiency in prototyping and developing deep learning models, making it a preferred choice, especially for beginners and for projects requiring rapid development.

After setting the variables for the environment, the next step is to install dependencies. To use Gemma, KerasNLP is the dependency used. KerasNLP is a collection of natural language processing (NLP) models implemented in Keras and runnable on JAX, PyTorch, and TensorFlow.

All breaking changes are listed here: https://github.com/keras-team/keras/issues/18467

You can use this migration guide to identify and fix each of these issues (and further, making your code run on JAX or PyTorch): https://keras.io/guides/migrating_to_keras_3/

I'm guessing the "real" keras code is coming from the keras repository. Is that a correct assumption? How does that version of Keras get there? If I wanted to write my own activation layer next to ELU, where exactly would I do that?

Other popular machine learning tools include PyTorch, Keras, and Scikit-learn. PyTorch is an open-source machine learning library developed by Facebook that is known for its ease of use and flexibility. Keras is a high-level neural networks API that is written in Python and is known for its simplicity. Scikit-learn is a machine learning library for Python that is used for data analysis and data mining tasks.

Click to Learn more...

I'm not aware of anything off-the-shelf, but if you have sufficient programming experience, one way to do this would be to build a large dataset of reference images and pictures and use something like keras to train a convolutional neural network on them.