Skip to content

dinov3

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

dinov3

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
Dockerfile
Dockerfile
 
 
README.md
README.md
 
 
config.yaml
config.yaml
 
 
test.sh
test.sh
 
 
test_dinov3.py
test_dinov3.py
 
 

README.md

DINOv3

Meta's DINOv3 vision foundation model for feature extraction.

Requirements

HuggingFace Token Required - DINOv3 models are gated and require authentication.

  1. Accept license at https://huggingface.co/facebook/dinov3-vits16-pretrain-lvd1689m (click on the Files and versions tab)
  2. Get token from https://huggingface.co/settings/tokens
  3. Update packages/vision/dinov3/config.yaml:
environment_variables:
- "HF_TOKEN=hf_your_actual_token_here"

Build and Run

ryzers build dinov3
ryzers run

Model

By default test.py uses facebook/dinov3-vits16-pretrain-lvd1689m (21.6M parameters - smallest variant).

Find list of models on Hugging Face, e.g:

  • facebook/dinov3-vitb16-pretrain-lvd1689m (85.7M)
  • facebook/dinov3-vitl16-pretrain-lvd1689m (0.3B)
  • facebook/dinov3-vit7b16-pretrain-lvd1689m (7B)

References

Copyright(C) 2025 Advanced Micro Devices, Inc. All rights reserved.