NVIDIA AI Foundry builds custom supermodels for enterprises using Llama 3.1 AI models.

From NVIDIA News: 2024-07-23 11:15:00

NVIDIA AI Foundry introduces a service allowing enterprises and nations to build custom “supermodels” using newly released Llama 3.1 models. Accenture is the first to use this service, creating custom models for generative AI applications. These supermodels can be trained with proprietary data and synthetic data generated by Llama 3.1 and NVIDIA Nemotron models.

NVIDIA AI Foundry offers an end-to-end service for building custom supermodels, combining NVIDIA software, infrastructure, and expertise with open community models and support from the NVIDIA AI ecosystem. Enterprises can create custom models using Llama 3.1 and NVIDIA NeMo platform, then deploy them using NVIDIA NIM inference microservices on their preferred cloud platforms and NVIDIA-Certified Systems™.

Enterprises can use Llama 3.1 and Nemotron-4 models together to generate synthetic data for enhanced model accuracy. NVIDIA and Meta provide a recipe for distillation to create smaller custom Llama models for a wider range of accelerated infrastructure, like AI workstations and laptops. Companies in various industries like healthcare, energy, and transportation are already benefiting from NVIDIA NIM microservices for Llama.

New NeMo Retriever NIM microservices improve response accuracy when deploying custom Llama supermodels in production. Paired with NIM microservices for Llama 3.1 405B models, NeMo Retriever enhances text question and answer retrieval accuracy. NVIDIA’s ecosystem of partners can integrate these microservices to enhance generative AI for over 5 million developers and 19,000 startups in the NVIDIA community.



Read more at NVIDIA News: NVIDIA AI Foundry Builds Custom Llama 3.1 Generative AI Models for the World’s Enterprises