Now loading...
Hugging Face, the popular platform for open-source AI models, has expanded its collaboration with Google Cloud to empower businesses in developing customized AI solutions using accessible open models. This strategic alliance aims to streamline the process of integrating and deploying AI technologies across various infrastructures.
Jeff Boudier, a key figure at Hugging Face, emphasized the significance of this move, stating that Google’s pioneering work in areas like transformers and the Gemma series has paved the way for broader adoption. He envisions a world where organizations of all sizes can tailor their own AI systems freely, and this partnership with Google Cloud brings that goal closer to reality.
Ryan J. Salva, Senior Director of Product Management at Google Cloud, highlighted Hugging Face’s role in democratizing AI by providing access to over two million open models, with Google contributing more than a thousand of its own. Together, they intend to position Google Cloud as the premier destination for constructing applications with these open resources.
For users of Google Cloud, Hugging Face’s open models are already embedded in flagship services such as Vertex AI, where top models can be launched quickly through the Model Garden interface. Those seeking finer control over their setups can access similar libraries in GKE for AI and machine learning workloads or use pre-built environments curated by Hugging Face. Additionally, serverless deployments for AI inference are possible via Cloud Run with GPU support, all designed to maximize the strengths of Google Cloud’s offerings while providing flexibility.
The partnership underscores a commitment to creating fluid user experiences that capitalize on each platform’s unique features, ensuring customers have ample options for their AI needs.
Over the past three years, Google Cloud users’ engagement with Hugging Face has surged tenfold, resulting in monthly downloads of tens of petabytes of models and billions of API requests. To enhance this momentum, the two companies are developing a dedicated content delivery network gateway for Hugging Face’s repositories. This system combines Hugging Face’s optimized storage and transfer tools with Google Cloud’s robust networking and storage infrastructure.
The gateway will store Hugging Face models and datasets closer to Google Cloud environments, slashing download times and bolstering the reliability of the AI model ecosystem. Benefits will extend to users on Vertex AI, GKE, Cloud Run, or even custom virtual machine setups in Compute Engine, leading to quicker response times and easier compliance management.
On the Hugging Face side, the deepened ties will infuse Google Cloud’s performance and efficiency into services like Inference Endpoints, which allow seamless model deployment with minimal effort. Users can anticipate expanded instance options and reduced costs as a direct outcome of this integration.
The collaboration ensures that innovations from joint product and engineering efforts reach Hugging Face’s community of ten million AI developers without friction. Deploying from a model page to Vertex AI’s Model Garden or GKE will be simpler, and handling private models in enterprise settings will match the simplicity of public ones.
Google’s Tensor Processing Units, now in their seventh iteration and renowned for AI acceleration, will become more accessible for open model development. Hugging Face’s libraries will natively support current and upcoming TPUs, making them as uncomplicated to manage as GPUs.
Security receives a major boost as well, with Hugging Face adopting Google Cloud’s advanced tools to safeguard its vast collection of open models. Drawing on technologies from VirusTotal, Google Threat Intelligence, and Mandiant, this initiative will protect models, datasets, and interactive Spaces against threats, enhancing daily usability on the Hugging Face Hub.
Ultimately, this alliance seeks to foster an era where every organization can craft and host its proprietary AI on open foundations within protected environments, maintaining complete oversight. By uniting forces on platforms like Vertex AI Model Garden, Google Kubernetes Engine, Cloud Run, and Hugging Face’s Inference Endpoints, they’re accelerating this open AI landscape.
