https://github.com/crcrpar/chainer-VAE

DEPRECATED!!! the same as official VAE example, but using Trainer in this repo
https://github.com/crcrpar/chainer-VAE

chainer deeplearning python

Last synced: 9 days ago
JSON representation

DEPRECATED!!! the same as official VAE example, but using Trainer in this repo

• Host: GitHub
• URL: https://github.com/crcrpar/chainer-VAE
• Owner: crcrpar
• Archived: true
• Created: 2016-11-08T04:38:50.000Z (about 8 years ago)
• Default Branch: master
• Last Pushed: 2018-02-15T07:34:16.000Z (over 6 years ago)
• Last Synced: 2024-08-02T12:21:58.936Z (3 months ago)
• Topics: chainer, deeplearning, python
• Language: Python
• Homepage:
• Size: 11 MB
• Stars: 30
• Watchers: 3
• Forks: 16
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# CAUTION

Now that chainer official implementation is updated. Please refer it.

# chainer-VAE

implementation of https://github.com/pfnet/chainer/tree/master/examples/vae using Trainer.

result/loss.png was compiled by using [Viz.js](https://mdaines.github.io/viz.js/).

If you use homebrew and wanna dump image file, `brew install graphviz`, then `dot -T   -o `.

## beta-VAE

As you can see in http://openreview.net/forum?id=Sy2fzU9gl, you can impose stronger regularization on latent space when define model via `net.VAE(C=beta)`, where, beta > 1.

### Execute `train_vae.py`

#### Argument Parser

- resume: resume from snapshot. Basically, snapshots are `chainer-VAE/result/snapshot_epoch_{epoch}`

- interval: save images & snapshot intervals, default is `5`