| ---
|
| license: apache-2.0
|
| tags:
|
| - base
|
| - concat-mlp
|
| - constant-warmup
|
| - flash
|
| - generation
|
| - groupnorm
|
| - mish
|
| - mobilevit
|
| - sgd
|
| - trunc-normal
|
| ---
|
|
|
| # train.py
|
|
|
| ## Model Overview
|
|
|
| A **base**-scale implementation of the **mobilevit** architecture, built for **generation** tasks.
|
|
|
| ## Architecture
|
|
|
| - **Architecture**: mobilevit
|
| - **Scale**: base
|
| - **Attention**: flash
|
| - **Fusion strategy**: concat mlp
|
| - **Task head**: generation
|
| - **Activation**: mish
|
| - **Normalization**: groupnorm
|
| - **Initialization**: trunc normal
|
|
|
| ## Training
|
|
|
| - **Optimizer**: sgd
|
| - **LR scheduler**: constant warmup
|
|
|
| ## Files
|
|
|
| - `train.py` — main artifact of this repository
|
|
|
| ## License
|
|
|
| See the license field above.
|
|
|
