| ---
|
| license: bsd-3-clause
|
| tags:
|
| - adamw
|
| - co-attention
|
| - contrastive
|
| - efficientformer
|
| - gelu
|
| - instancenorm
|
| - linear-warmup
|
| - small
|
| - sparse
|
| - xavier
|
| ---
|
|
|
| # predict.py
|
|
|
| ## Model Overview
|
|
|
| A **small**-scale implementation of the **efficientformer** architecture, built for **contrastive** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: efficientformer
|
| - **Scale**: small
|
| - **Attention**: sparse
|
| - **Fusion strategy**: co attention
|
| - **Task head**: contrastive
|
| - **Activation**: gelu
|
| - **Normalization**: instancenorm
|
| - **Initialization**: xavier
|
|
|
| ## Training
|
|
|
| - **Optimizer**: adamw
|
| - **LR scheduler**: linear warmup
|
|
|
| ## Files
|
|
|
| - `predict.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
