https://github.com/nextml-code/template-nvae

Simple NVAE implementation intended for effective reuse and learning purposes
https://github.com/nextml-code/template-nvae

generative-model nvae pytorch

Last synced: 1 day ago
JSON representation

Simple NVAE implementation intended for effective reuse and learning purposes

• Host: GitHub
• URL: https://github.com/nextml-code/template-nvae
• Owner: nextml-code
• Created: 2020-08-27T13:06:32.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2020-10-29T00:37:44.000Z (about 4 years ago)
• Last Synced: 2023-03-05T13:34:09.156Z (almost 2 years ago)
• Topics: generative-model, nvae, pytorch
• Language: Python
• Homepage:
• Size: 7.22 MB
• Stars: 16
• Watchers: 4
• Forks: 1
• Open Issues: 3
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Nouveau Variational Autoencoder (NVAE)

Simple [NVAE](https://arxiv.org/abs/2007.03898) implementation to be used as 

template for training on other datasets. Intended for effective reuse and 

learning purposes. Will be improved on continually.

Images generated with prior standard deviation (aka temperature) from 0.1

to 1.1 (top to bottom) after first updating batch norms with samples

generated with prior standard deviation 1.0.

![Generated images with different temperatures](assets/prior_std100.png)

## Installation

```

virtualenv venv --python python3.8

source venv/bin/activate

pip install -r requirements.txt

```

## Prepare

Downloads and prepares dataset in a folder like in a normal setting.

```

guild run prepare

```

_You may get an error saying "bad zip file". This means that the limit on

downloading the celeba dataset has been reached._

## Training

```

guild run train

guild run retrain model=

guild run generate model=

```

These are some of the most relevant guild commands:

```

guild tensorboard 1

guild runs

```

See the [guild docs](https://my.guild.ai/docs) for more information.