Vertex AI Model Garden
Vertex AI Model Garden is a collection of pre-built foundation models, task-specific models, and Google ML APIs.
Today, we’re diving into Google’s Vertex AI Model Garden. It’s like a playground for AI models, with an impressive collection of pre-built foundation models, task-specific models, and Google ML APIs.
As of this article, 127+ models are available. Including many open-source and third-party models like Llama 2, Stable Diffusion, and Falcon.
The available models are a steadily growing list, including soon also Claude 2.
Google Model Garden provides access to the model in various ways. Depending on which model you choose, you see different options.
- Generative AI Studio
Google’s Generative AI Models are supported in Google’s Generative AI Studio. A UI to interactively work with Gen AI models. This includes PaLM for Text, Imagen for Images, and Chirp for Speech. This is great for initial POCs.
Google’s Models available via Generative AI Studio can also be accessed via API to be integrated into applications and put into production.
The open-source models 81+ in the Vertex AI Model Garden are available via Jupiter Notebook. The notebooks help you to start the model training/tuning and deployment quickly. All you need is in the notebook, ready to be used or adapted to your needs.
- One-click deployments
As the name suggests, you can, with one click, deploy models like Llama 2 and Falcon behind a Vertex AI Endpoint ready for you to use. Google automatically uses the proper hardware and serving-optimized image.
A few models are available as Vertex AI Pipelines. We should see more open-source models made available like this.
Each model has a model card that provides detailed information about the model, license, sizes (e.g., 7B, 13B, 70B), and documentation.
It also includes deploying open-source models (one-click deployment) or opening them as a notebook. For Google’s Generative Models, you can open the Generative AI Studio and see code examples on integrating…