dongjunlee / vae-tensorflow

TensorFlow implementation of Auto-Encoding Variational Bayes.

Variational Autoencoder

TensorFlow implementation of Auto-Encoding Variational Bayes.

Requirements

• Python 3.6
• TensorFlow >= 1.4
• hb-config (Singleton Config)
• requests
• Slack Incoming Webhook URL
• Matplotlib

Project Structure

init Project by hb-base

.
├── config                  # Config files (.yml, .json) using with hb-config
├── data                    # dataset path
├── variational_autoencoder     # VAE architecture graphs (from input to logits)
    └── __init__.py             # Graph logic
├── data_loader.py          # download data -> generate_batch (using Dataset)
├── main.py                 # define experiment_fn
└── model.py                # define EstimatorSpec

Reference : hb-config, Dataset, experiments_fn, EstimatorSpec

Config

Can control all Experimental environment.

example: mnist.yml

model:
  batch_size: 32
  z_dim: 20
  n_output: 784

  encoder_h1: 512
  encoder_h2: 256
  encoder_h3: 128

  decoder_h1: 128
  decoder_h2: 256
  decoder_h3: 512

train:
  learning_rate: 0.00001
  optimizer: 'Adam'                # Adagrad, Adam, Ftrl, Momentum, RMSProp, SGD

  train_steps: 200000
  model_dir: 'logs/mnist'

  save_checkpoints_steps: 1000
  check_hook_n_iter: 1000
  min_eval_frequency: 10

  print_verbose: True
  debug: False

slack:
  webhook_url: ""                   # after training notify you using slack-webhook

• debug mode : using tfdbg

Usage

Install requirements.

pip install -r requirements.txt

Then, start training model

python main.py --config mnist

After training, generate image from latent vector.

python generate.py --config mnist --batch_size 100

Experiments modes

✅ : Working
◽ : Not tested yet.

• ✅ evaluate : Evaluate on the evaluation data.
• ◽ extend_train_hooks : Extends the hooks for training.
• ◽ reset_export_strategies : Resets the export strategies with the new_export_strategies.
• ◽ run_std_server : Starts a TensorFlow server and joins the serving thread.
• ◽ test : Tests training, evaluating and exporting the estimator for a single step.
• ✅ train : Fit the estimator using the training data.
• ✅ train_and_evaluate : Interleaves training and evaluation.

---

Tensorboar

tensorboard --logdir logs

Result

• Generate Mnist image (Config: mnist.yml)

Reference

• hb-research/notes - Auto-Encoding Variational Bayes
• Paper - Auto-Encoding Variational Bayes
• shaohua0116/VAE-Tensorflow

Author

Dongjun Lee ([email protected])