model_685842618_dino_base.py
Model Overview
A base-scale implementation of the dino architecture, built for multitask tasks.
Architecture
- Architecture: dino
- Scale: base
- Attention: multi query
- Fusion strategy: co attention
- Task head: multitask
- Activation: approx gelu
- Normalization: rmsnorm
- Initialization: xavier uniform
Training
- Optimizer: adafactor
- LR scheduler: onecycle
Files
model_685842618_dino_base.py — main artifact of this repository
License
See the license field above.