| ---
|
| license: cc-by-4.0
|
| tags:
|
| - blip
|
| - contrastive
|
| - gated-fusion
|
| - gelu-tanh
|
| - instancenorm
|
| - lamb
|
| - linear
|
| - orthogonal
|
| - step
|
| - xlarge
|
| ---
|
|
|
| # main.py
|
|
|
| ## Model Overview
|
|
|
| A **xlarge**-scale implementation of the **blip** architecture, built for **contrastive** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: blip
|
| - **Scale**: xlarge
|
| - **Attention**: linear
|
| - **Fusion strategy**: gated fusion
|
| - **Task head**: contrastive
|
| - **Activation**: gelu tanh
|
| - **Normalization**: instancenorm
|
| - **Initialization**: orthogonal
|
|
|
| ## Training
|
|
|
| - **Optimizer**: lamb
|
| - **LR scheduler**: step
|
|
|
| ## Files
|
|
|
| - `main.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
