Nvidia bets at least $500 million on Inception's diffusion models to strengthen its open-source Nemotron lineup
NVIDIA has released Nemotron-Labs-Sigma, a collection of 14 open-source language model checkpoints available in 3B and 8B parameter sizes. By utilizing continuous diffusion to denoise text within an embedding space, these models can decode information up to 4.20 times faster than traditional autoregressive methods. This approach aims to improve inference efficiency through few-step sampling techniques.
Covered by 2 sources
- TThe Decoder↗Maximilian Schreiner1d ago
- MMarkTechPost↗Asif Razzaq1d ago