Loading...
Loading...
Browse all stories on DeepNewz
VisitWhat will be recognized as the primary benefit of nGPT by mid-2025?
Training Speed • 25%
Stability • 25%
Energy Efficiency • 25%
Other • 25%
Industry analysis reports and AI research papers
NVIDIA Unveils Normalized Transformer (nGPT) for 4-20x Faster Training and Improved Stability of Large Language Models
Oct 19, 2024, 02:20 PM
NVIDIA researchers have introduced a new architecture called the Normalized Transformer (nGPT), which utilizes a hypersphere-based design to enhance the training speed and stability of large language models (LLMs). The nGPT architecture reportedly achieves training speeds that are 4 to 20 times faster than previous models. This advancement is expected to significantly improve convergence speed in transformer models, as noted by various experts in the field. The introduction of nGPT marks a notable development in the ongoing evolution of AI and machine learning technologies.
View original story
Content Creation • 25%
Customer Support • 25%
Data Analysis • 25%
Entertainment • 25%
Customer service automation • 25%
Content creation • 25%
Data analysis • 25%
Other applications • 25%
Chatbots • 25%
Content Creation • 25%
Data Analysis • 25%
Other • 25%
Only US • 25%
North America • 25%
Europe • 25%
Global • 25%
Genetic Variation Analysis • 25%
Recombination Event Detection • 25%
Phylogenetic Relationship Inference • 25%
Conservation Measurement • 25%
Healthcare • 25%
Financial Services • 25%
Education • 25%
Entertainment • 25%
Speed of simulation • 25%
Ease of integration • 25%
Realism of simulations • 25%
Other • 25%
Gaming • 25%
Virtual Reality • 25%
Simulation Training • 25%
Other • 25%
Virtual Reality • 25%
Film Production • 25%
Military Training • 25%
Other • 25%
Yes • 50%
No • 50%
Chatbots • 25%
Content creation • 25%
Customer support • 25%
Educational tools • 25%
Behavioral Simulation • 25%
Customer Service • 25%
Content Creation • 25%
Data Analysis • 25%
Chatbots and customer service • 25%
Content creation • 25%
Data analysis and insights • 25%
Other • 25%
No • 50%
Yes • 50%
CVPR 2024 • 25%
NeurIPS 2024 • 25%
Other • 25%
ICML 2024 • 25%