model_691747019_dino_large.py

Model Overview

A large-scale implementation of the dino architecture, built for contrastive tasks.

Architecture

• Architecture: dino
• Scale: large
• Attention: dilated
• Fusion strategy: gated fusion
• Task head: contrastive
• Activation: gelu
• Normalization: scalenorm
• Initialization: kaiming

Training

• Optimizer: sgd
• LR scheduler: constant warmup

Files

• model_691747019_dino_large.py — main artifact of this repository

License

See the license field above.