metadata
license: mit
tags:
- approx-gelu
- base
- bilinear
- contrastive
- dilated
- kaiming-normal
- layernorm
- novograd
- polynomial
- poolformer
model_427145758_poolformer_base.py
Model Overview
A base-scale implementation of the poolformer architecture, built for contrastive tasks.
Architecture
- Architecture: poolformer
- Scale: base
- Attention: dilated
- Fusion strategy: bilinear
- Task head: contrastive
- Activation: approx gelu
- Normalization: layernorm
- Initialization: kaiming normal
Training
- Optimizer: novograd
- LR scheduler: polynomial
Files
model_427145758_poolformer_base.py— main artifact of this repository
License
See the license field above.