| ---
|
| license: mit
|
| tags:
|
| - batchnorm
|
| - cnn-transformer
|
| - concat-mlp
|
| - constant-warmup
|
| - linear
|
| - nano
|
| - retrieval
|
| - rmsprop
|
| - swish
|
| - xavier-uniform
|
| ---
|
|
|
| # train.py
|
|
|
| ## Model Overview
|
|
|
| A **nano**-scale implementation of the **cnn transformer** architecture, built for **retrieval** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: cnn transformer
|
| - **Scale**: nano
|
| - **Attention**: linear
|
| - **Fusion strategy**: concat mlp
|
| - **Task head**: retrieval
|
| - **Activation**: swish
|
| - **Normalization**: batchnorm
|
| - **Initialization**: xavier uniform
|
|
|
| ## Training
|
|
|
| - **Optimizer**: rmsprop
|
| - **LR scheduler**: constant warmup
|
|
|
| ## Files
|
|
|
| - `train.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
