# Training with Lightly Filtered Web Data

A 2-billion-parameter vision Transformer model trained using the DINOv2 self-supervised learning framework on lightly filtered web-scale image data (without language supervision).