metadata
license: apache-2.0
tags:
- approx-gelu
- blip
- contrastive
- lamb
- scalenorm
- sliding-window
- step
- tensor-fusion
- tiny
- xavier
model.py
Model Overview
A tiny-scale implementation of the blip architecture, built for contrastive tasks.
Architecture
- Architecture: blip
- Scale: tiny
- Attention: sliding window
- Fusion strategy: tensor fusion
- Task head: contrastive
- Activation: approx gelu
- Normalization: scalenorm
- Initialization: xavier
Training
- Optimizer: lamb
- LR scheduler: step
Files
model.py— main artifact of this repository
License
See the license field above.