run.py
Model Overview
A xlarge-scale implementation of the dino architecture, built for matching tasks.
Architecture
- Architecture: dino
- Scale: xlarge
- Attention: standard
- Fusion strategy: tucker
- Task head: matching
- Activation: gelu
- Normalization: scalenorm
- Initialization: xavier
Training
- Optimizer: lion
- LR scheduler: constant warmup
Files
run.py — main artifact of this repository
License
See the license field above.
