train.py
Model Overview
A small-scale implementation of the dino architecture, built for retrieval tasks.
Architecture
- Architecture: dino
- Scale: small
- Attention: grouped query
- Fusion strategy: bilinear
- Task head: retrieval
- Activation: gelu
- Normalization: layernorm
- Initialization: trunc normal
Training
- Optimizer: adam
- LR scheduler: linear warmup
Files
train.py — main artifact of this repository
License
See the license field above.