Nvidia corporation (20250078489). FULLY ATTENTIONAL NETWORKS WITH SELF-EMERGING TOKEN LABELING
FULLY ATTENTIONAL NETWORKS WITH SELF-EMERGING TOKEN LABELING
Organization Name
Inventor(s)
Bingyin Zhao of Central SC (US)
Jose Manuel Alvarez Lopez of Mountain View CA (US)
Anima Anandkumar of Pasadena CA (US)
Shi Yi Lan of San Jose CA (US)
Zhiding Yu of Santa Clara CA (US)
FULLY ATTENTIONAL NETWORKS WITH SELF-EMERGING TOKEN LABELING
This abstract first appeared for US patent application 20250078489 titled 'FULLY ATTENTIONAL NETWORKS WITH SELF-EMERGING TOKEN LABELING
Original Abstract Submitted
one embodiment of the present invention sets forth a technique for training an image classifier. the technique includes training a first vision transformer model to generate patch labels for corresponding images patches of images, converting the patch labels to token labels, and training a second vision transformer model to classify images based on the token labels.