metadata
license: apache-2.0
tags:
- approx-gelu
- groupnorm
- low-rank
- matching
- novograd
- orthogonal
- polynomial
- poolformer
- sliding-window
- tiny
model.py
Model Overview
A tiny-scale implementation of the poolformer architecture, built for matching tasks.
Architecture
- Architecture: poolformer
- Scale: tiny
- Attention: sliding window
- Fusion strategy: low rank
- Task head: matching
- Activation: approx gelu
- Normalization: groupnorm
- Initialization: orthogonal
Training
- Optimizer: novograd
- LR scheduler: polynomial
Files
model.py— main artifact of this repository
License
See the license field above.