Oracle Cloud Infrastructure’s Generative AI service is the latest AI service in the OCI AI catalog and it offers advanced large language models (LLMs) tailored for various text generation needs. You can explore and experiment with the pre-trained models in the playground, or develop and deploy your own customized models using your data on specialized AI clusters.
Playground
The playground is a web interface in the Oracle Cloud Console that lets you explore hosted pre-trained and custom models without any coding. Use the playground to test use cases and refine prompts and parameters.
You simply add your input, select the desired model and press….Generate 🙂
This allows us to experiment, play with the different parameters, try prompt variations and get a feel for which model matches the needs.
There are 3 features/capabilities that we can use (Generation/Summarization/embeddings)
Generation
Give instructions to generate text or extract information from your text.
The following pre-trained foundational text generation models are available in Generative AI:
cohere.commandcohere.command-lightmeta.llama-2-70b-chat
Cohere.command
A generation model with 50 billion parameters with basic world knowledge. Ideal for tasks like text extraction, sentiment analysis, and drafting marketing content, emails, blog posts, and product descriptions.
Cohere.command-light
A fast, lightweight generation model for tasks needing basic world knowledge and simple instructions. Best for speed and cost efficiency. Clear, specific prompts improve performance.
Meta.llama-2-70b-chat
A high-performance model with 70 billion parameters and broad world knowledge. Suitable for brainstorming, text extraction, sentiment analysis, and creating marketing content, emails, blog posts, and product descriptions.
Summarization
Summarize text with your instructed format, length, and tone.
OCI Generative AI offers a version of the cohere.command model specifically for text summarization. It’s just like the standard text generation model but fine-tuned with parameters for summarizing
Embedding
Convert text to vector embeddings to use in applications for semantic searches, text classification, or text clustering.
OCI Generative AI embedding models turn any phrase, sentence, or paragraph you input into an array of numbers.
These embeddings are great for identifying phrases with similar context or category. Usually stored in a vector database, embeddings are primarily used for semantic searches, focusing on the meaning of the text rather than just keywords.
In the below example I append a completely unrelated sentence, which will reflect in the generated embeddings.
And here you can see a plotted graph with a perfect example of, “the one not like the others”.
All of this information and much more can be found in the documentation here.