---
license: cc-by-4.0
tags:
- contrastive
- grouped-query
- groupnorm
- linear-warmup
- nano
- novograd
- orthogonal
- swish
- tucker
- vit
---

# main.py

## Model Overview

A **nano**-scale implementation of the **vit** architecture, built for **contrastive** tasks.

## Architecture

- **Architecture**: vit
- **Scale**: nano
- **Attention**: grouped query
- **Fusion strategy**: tucker
- **Task head**: contrastive
- **Activation**: swish
- **Normalization**: groupnorm
- **Initialization**: orthogonal

## Training

- **Optimizer**: novograd
- **LR scheduler**: linear warmup

## Files

- `main.py` — main artifact of this repository

## License

See the license field above.