| ---
|
| license: bsd-3-clause
|
| tags:
|
| - albef
|
| - bilinear
|
| - cosine
|
| - generation
|
| - giant
|
| - multi-query
|
| - rmsprop
|
| - scalenorm
|
| - swish
|
| - trunc-normal
|
| ---
|
|
|
| # train.py
|
|
|
| ## Model Overview
|
|
|
| A **giant**-scale implementation of the **albef** architecture, built for **generation** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: albef
|
| - **Scale**: giant
|
| - **Attention**: multi query
|
| - **Fusion strategy**: bilinear
|
| - **Task head**: generation
|
| - **Activation**: swish
|
| - **Normalization**: scalenorm
|
| - **Initialization**: trunc normal
|
|
|
| ## Training
|
|
|
| - **Optimizer**: rmsprop
|
| - **LR scheduler**: cosine
|
|
|
| ## Files
|
|
|
| - `train.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
