| ---
|
| license: apache-2.0
|
| tags:
|
| - cross-attention
|
| - gelu
|
| - groupnorm
|
| - huge
|
| - linear
|
| - mobilevit
|
| - onecycle
|
| - retrieval
|
| - rmsprop
|
| - trunc-normal
|
| ---
|
|
|
| # model.py
|
|
|
| ## Model Overview
|
|
|
| A **huge**-scale implementation of the **mobilevit** architecture, built for **retrieval** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: mobilevit
|
| - **Scale**: huge
|
| - **Attention**: linear
|
| - **Fusion strategy**: cross attention
|
| - **Task head**: retrieval
|
| - **Activation**: gelu
|
| - **Normalization**: groupnorm
|
| - **Initialization**: trunc normal
|
|
|
| ## Training
|
|
|
| - **Optimizer**: rmsprop
|
| - **LR scheduler**: onecycle
|
|
|
| ## Files
|
|
|
| - `model.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
