Support for embeddings

[mqtik]

Problem description

Currently, the library supports text generation with models like Llama but lacks functionality for generating text embeddings. Text embeddings are crucial for features such as semantic search and content recommendations.

Proposed solution

Support embeddings. This could involve integrating models capable of producing embeddings, like BERT or Universal Sentence Encoder, and extending the API to include methods for obtaining embeddings directly from input text.

Alternative solutions

While other libraries like react-native-transformers or tensorflow.js offer embedding capabilities, integrating them may lead to compatibility issues or increased complexity within the development environment

Benefits to React Native ExecuTorch

Implementing text embedding capabilities would enhance the library's versatility, enabling a wider range of AI-driven features within React Native applications

Additional context

No response