| ---
|
| license: apache-2.0
|
| tags:
|
| - base
|
| - batchnorm
|
| - blip
|
| - exponential
|
| - generation
|
| - lamb
|
| - orthogonal
|
| - sparse
|
| - swish
|
| - tensor-fusion
|
| ---
|
|
|
| # eval.py
|
|
|
| ## Model Overview
|
|
|
| A **base**-scale implementation of the **blip** architecture, built for **generation** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: blip
|
| - **Scale**: base
|
| - **Attention**: sparse
|
| - **Fusion strategy**: tensor fusion
|
| - **Task head**: generation
|
| - **Activation**: swish
|
| - **Normalization**: batchnorm
|
| - **Initialization**: orthogonal
|
|
|
| ## Training
|
|
|
| - **Optimizer**: lamb
|
| - **LR scheduler**: exponential
|
|
|
| ## Files
|
|
|
| - `eval.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
