Nvidia is expanding its Inference Microservices offering to include training capabilities for 3D worlds. In addition, Nvidia will support Hugging Face users in faster training of AI models.
Hugging Face users will have access to new inferencing-as-a-service services running on Nvidia’s DGX Cloud. Users of the machine learning platform can claim the cloud for faster performance. The new capability should mainly help developers figure out the right AI model for business needs.
With Nvidia Inference Microservices (NIM), an enterprise offering is envisioned for faster and easier deployment of AI models. Through the offering, developers can additionally train AI models with enterprise data.
3D training
One thing NIM is now investing in is training three-dimensional AI. This includes the new Fast Voxel Database (fVDB) NIM service. The fVDB deep-learning framework itself needs a brief introduction. The framework is an extension of OpenVDB and allows developers to create virtual representations of places in the real world. It has four times the spatial scale of previous frameworks, with three and a half times better performance. Developers can access a large number of datasets through fVDB.
Use cases for the framework are robots or self-driving cars. Such things need a virtual copy of the real world to operate safely.
In addition, Nvidia announced the preview of USD Code, USD Search and USD Validate. The basis of these services is the Universal Scene Description format. The Code service works for generating Python. The Search offering allows developers to search through OpenUSD 3D data with natural language queries. The latter service assures developers that the uploaded files are also compatible with OpenUSD release versions.
“The generative AI boom for heavy industry has arrived,” said Rev Lebaredian, vice president of Omniverse and simulation technology at Nvidia. “Until recently, digital worlds have been primarily used by creative industries; now, with the enhancements and accessibility Nvidia NIM microservices are bringing to OpenUSD, industries of all kinds can build physically based virtual worlds and digital twins to drive innovation while preparing for the next wave of AI: robotics.”
Also read: Nvidia is currently the most popular kid in class