Today we are excited to announce the launch of Deploy for Google Cloud, a new integration of Hugging Face Hub, Google Cloud, to easily deploy thousands of foundation models to the Google Cloud using Vertex AI or Google Kubernetes Engine (GKE). With deployment on Google Cloud, Open Models can easily be deployed as an API endpoint within your Google Cloud account, either by hugging Google model cards or via Vertex Model Garden, a single-player Google Cloud for discovering, customizing and deploying a variety of models from Google and Google partners. From today, we are enabling the most popular open models to hug faces for inference that drives the inference of text generation, a production solution.
Deploying it to Google Cloud allows developers to build production-ready, generic AI applications without directly managing infrastructure and servers within a secure Google Cloud environment.
AI Builder Collaboration
This new experience expands the strategic partnership announced earlier this year to simplify access and deployment of Google’s open-generated AI models for its customers. One of the main issues faced by developers and organizations is the time and resources needed to safely and reliably deploy the model. Deployment on Google Cloud provides a simple, managed solution to these challenges, providing dedicated configurations and assets to hug face models. Creating a production-enabled endpoint in Google Cloud’s Vertex AI is a simple click-through experience.
“The Vertex AI model garden integration seamlessly creates discovery and deploy open models with Vertex AI and GKE, whether you’re going on a hub or starting directly on the Google Cloud console.” “I can’t wait to see what Google developers embrace and build Face Models.”
How it works – From the hub
It’s very easy to deploy a face model that hugs to Google Cloud. Below are step-by-step instructions on how to deploy Zephyr Gemma. Starting today, all models with the “Text Generation Inference” tag are supported.
Open the (Expand) menu and select (Google Cloud). This will be straightened directly to the Google Cloud Console. Here you can expand Zephyr Gemma with vertex AI or GKE by clicking 1.
Once in Vertex Model Garden, you can choose Vertical AI or GKE as your deployment. Vertex AI allows you to expand the model by clicking “Expand” one click. For GKE, you can follow instructions and manifest templates on how to deploy the model to a new or running Kubernetes cluster.
How it works – From the Vertex Model Garden
Vertex Model Garden is where Google developers can find ready-to-use models for generated AI projects. Starting today, Vertex Model Garden offers a new experience for easily deploying the most popular open LLM available on Face!
You can find a new “Deploy from Hug” option within Google Vertex AI Model Garden. This allows you to directly search and deploy the face model of your hug within the Google Cloud Console.
Click “Expand from Hug” to display a form that lets you quickly search for model IDs. Hundreds of open LLMs that are most popular with embracing faces are available in ready-to-use, tested hardware configurations.
Once you find the model you want to deploy, select it. Vertex AI prepares all the configuration needed to deploy the model to Vertex AI or GKE. You can also check that you have selected the right model by “hugging your face and looking.” If you are using a gate model, make sure to provide a hugging face access token so that you can approve the model’s download.
And that’s it! It’s only a few clicks to deploy a model like Zephyr Gemma directly and move it from Vertex Model Garden to your Google Cloud account.
I’ve just started
We are excited to work with Google Cloud to make AI more open and accessible. Deploying an open model to Google Cloud has never been easier, whether you start with a hugging facehub or a Google Cloud console. And we’re not going to stop there – look forward to it as it allows for more experiences to build AI using open models on Google Cloud.
