model_691991104_dino_large.py

Model Overview

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

Architecture

• Architecture: dino
• Scale: large
• Attention: dilated
• Fusion strategy: tensor fusion
• Task head: retrieval
• Activation: relu
• Normalization: batchnorm
• Initialization: orthogonal

Training

• Optimizer: adamw
• LR scheduler: cosine

Files

• model_691991104_dino_large.py — main artifact of this repository

License

See the license field above.