Today, GTC Paris is excited to announce its training cluster as a collaborative service with NVIDIA, making it easier to access large GPU clusters for research organizations around the world.
Make your GPU cluster accessible
Many gigawatt size GPU super cluster projects have been built and trained the next Gen AI model. This can sometimes seem to expand the calculation gap between “GPU poor” and “GPU rich”. But Hyperscalers, Regional and AI-Native cloud providers all expand their capabilities quickly, so the GPU is there.
How do you connect with researchers who need AI computing power? How can universities, national labs and businesses around the world be able to build their own models?
This is a hugging face and Nvidia is working on training clusters as a service, providing accessibility to GPU clusters with the flexibility to pay only for the duration of the training.
To get started, any of the 250,000 organizations hugging your face can request the GPU cluster size they need when they need it.
How it works
To get started, you can request a GPU cluster on your organization’s behalf with hf.co/training-cluster
Training as a Service cluster integrates key components from Nvidia and embraces your face.
Nvidia Cloud Partners offers the latest NVIDIA accelerated computing capabilities, such as the NVIDIA Hopper and Nvidia GB200 regional data centers. It’s easy to start running your training.
Once the GPU cluster request is accepted, Face and Nvidia embrace collaborates with the source, price, offering and setup of the GPU cluster by size, area, and duration requirements.
Clusters at work
Advances in rare genetic disease research using Tigem
The Teleson Institute of Genomics and Medical Institute – Tigem for Short is a research center dedicated to understanding the molecular mechanisms behind rare genetic diseases and developing new treatments. Training new AI models is a new pathway for drug relocation, predicting the effects of pathogenic mutants.
AI offers new ways to study the causes of rare genetic diseases and develop treatments, but our domain requires training of new models. Training clusters as a service made it easy to obtain the required GPU capacity at the right time
– Diego Di Bernardo, Coordinator of Tigem’s Genomic Medicine Program
Use numina for math to advance AI
Numina is a nonprofit organization that builds open source, open source, open data set AI for mathematical reasoning, and was awarded the 2024 Aimo Progress Award.
It tracks well the purpose of building open alternatives to the best closed source models, such as DeepMind’s alphaproof. Computing resources are today’s bottleneck – training clusters will become available as services.
– Yann Fleueau, co-founder of Project Numina
Advance materials science using mirror physics
Mirror Physics is a startup that creates frontier AI systems in chemistry and materials science.
Together with the MACE team, we are committed to pushing the limits of AI for chemistry. By training clusters as services, we produce high fidelity chemical models on an unprecedented scale. This will be an important step forward for the field.
– Miller CEO and founder Sam Walton Norwood
Promoting diversity in AI research
Training as a Service Cluster is a new collaboration that embraces faces and Nvidia to make AI researcher calculations easier to use.
Access to large-scale, high-performance computing is essential to building next-generation AI models across all domains and languages. Training clusters as a service removes barriers for researchers and businesses, disengages the ability to train the most advanced models and push the boundaries of what AI can do.
-Clément Delangue, co-founder and CEO of Hugging Face
Integrating DGX Cloud Lepton and Face training clusters as a service provides a seamless way for developers and researchers to access high-performance NVIDIA GPUs across a wide network of cloud providers. This collaboration makes it easier for AI researchers and organizations to use easy-to-hugging tools while expanding their AI training workloads.
-Alexis Byolin, Vice President of DGX Cloud at NVIDIA
Enable AI Builder in NVIDIA
We look forward to working with Nvidia to offer training clusters as a service to embrace facial organizations. You can start today with hf.co/training -cluster.
Today, at GTC Paris, Nvidia has announced many new contributions to hugging face users, from agents to robots!
