On This Page: [hide]
Google Cloud’s esteemed cloud unit has recently unveiled a significant collaboration with Hugging Face Inc., the operator of a widely-used platform for sharing open-source artificial intelligence models.
This strategic partnership positions Google Cloud as the “preferred destination” for Hugging Face’s AI training and inference workloads, fostering a deeper integration of their respective services.
Hugging Face’s Diverse Portfolio
Hugging Face, valued at $4.5 billion as of last August, operates a platform akin to GitHub, facilitating the hosting of open-source AI models and sharing associated files, including training datasets. Beyond hosting machine learning projects from various developers, Hugging Face has contributed to the field by creating two open-source language models through research partnerships.
The company monetizes its platform through a suite of paid cloud services, including AutoTrain, which streamlines the manual effort involved in training AI models, and Inference Endpoints for hosting pre-trained AI models. With this collaboration, Hugging Face aims to leverage Google Cloud’s Vertex AI product suite, which incorporates AI development tools and over 130 prepackaged foundational models. Developers will seamlessly access Vertex AI from the Hugging Face interface, enhancing the ease of AI development with just a few clicks.
The integration extends to Google Kubernetes Engine (GKE), Google’s managed Kubernetes service, allowing developers to run AI workloads, including open-source neural networks from Hugging Face, in the form of software containers. This integration facilitates the deployment of containers on Google’s cloud platform, leveraging AI-optimized chips for enhanced performance.
Hugging Face and Google Cloud Collaboration
Hugging Face users will also gain access to Google Cloud TPU v5e, the latest machine learning processor developed by Google internally. This chip promises a remarkable 250% improvement in performance per dollar compared to its predecessor, with reduced latency for running AI models efficiently.
Looking ahead, Hugging Face plans to support Google’s recently launched A3 VMs. These instances feature eight H100 graphics cards from Nvidia Corp., offering faster data exchange capabilities to enhance AI performance. Supported by two Intel Corp. processors and boasting two terabytes of memory, the A3 VMs represent a great infrastructure for cutting-edge AI applications.
Google Cloud’s Chief Executive, Thomas Kurian, highlighted the shared vision of making generative AI more accessible and impactful for developers. The partnership aims to provide Hugging Face developers with access to Google Cloud’s purpose-built AI platform, Vertex AI, coupled with a secure infrastructure to accelerate the next generation of AI services and applications.
In addition to the infrastructure enhancements, Hugging Face will integrate with Google Cloud Marketplace to simplify billing tasks for customers utilizing its paid AI development and hosting services. This integration will also extend to the Spaces portal, a platform used by open-source developers to share AI applications, further streamlining the user experience within the Hugging Face ecosystem.