---
license: mit
---
# pixel

A PyTorch-based Generative Adversarial Network (GAN) for training and generating pixel art images.

## Setup for model training

Git clone the pixel repo:
```
git clone https://github.com/mochiyaki/pixel
```

Get inside the cloned folder:
```
cd pixel
```

## Train your model

Start training with your dataset (in ./data/):
```
python trainer.py
```

![screenshot](https://raw.githubusercontent.com/mochiyaki/pixel/master/running.png)

* kick start the training process with defined script

![screenshot](https://raw.githubusercontent.com/mochiyaki/pixel/master/graph.png)

* check the training epoch (currently 50); see the example plot above

## Interact with the established model

When finished, check the model file (in ./models/) then run the inference:
```
python generator.py
```

![screenshot](https://raw.githubusercontent.com/mochiyaki/pixel/master/demo.gif)

* for technical details, please check it out from [this repo](https://github.com/mochiyaki/pixel)

## alternatively opt to run `ggc px` with `gguf-connector`