VAE (TensorFlow)

An implementation of VAE language generation model. Refer to the following paper for more details:

Bowman, S., Vilnis, L., Vinyals, O., Dai, A., Jozefowicz, R., and Bengio, S. Generating sentences from a continuous space. In Proceedings of the Twentieth Conference on Computational Natural Language Learning, 2016.

Require Packages

• cotk

• TensorFlow == 1.13.1

• TensorBoardX >= 1.4

Quick Start

• Downloading dataset and save it to ./data. (Dataset will be released soon.)

• Execute python run.py to train the model.

  • The default dataset is MSCOCO. You can use --dataset to specify other dataloader class.

  • It use gloves pretrained word vector by default setting. You can use --wvclass to specify wordvector class.

  • If you don’t have GPUs, you can add --cpu for switching to CPU, but it may cost very long time.

• You can view training process by tensorboard, the log is at ./tensorboard.

  • For example, tensorboard --logdir=./tensorboard. (You have to install tensorboard first.)

• After training, execute python run.py --mode test --restore best for test.

  • You can use --restore filename to specify checkpoints files, which are in ./model.

  • --restore last means last checkpoint, --restore best means best checkpoints on dev.

• Find results at ./output.

Arguments

usage: run.py [-h] [--name NAME] [--restore RESTORE] [--mode MODE]
              [--dataset DATASET] [--datapath DATAPATH] [--epoch EPOCH]
              [--wvclass WVCLASS] [--wvpath WVPATH] [--out_dir OUT_DIR]
              [--log_dir LOG_DIR] [--model_dir MODEL_DIR]
              [--cache_dir CACHE_DIR] [--cpu] [--debug] [--cache]


optional arguments:
  -h, --help            show this help message and exit
  --name NAME           The name of your model, used for variable scope and
                        tensorboard, etc. Default: runXXXXXX_XXXXXX
                        (initialized by current time)
  --restore RESTORE     Checkpoints name to load. "last" for last checkpoints,
                        "best" for best checkpoints on dev. Attention: "last"
                        and "best" wiil cause unexpected behaviour when run 2
                        models in the same dir at the same time. Default: None
                        (dont load anything)
  --mode MODE           "train" or "test". Default: train
  --dataset DATASET     Dataloader class. Default: MSCOCO
  --datapath DATAPATH   Directory for data set. Default: ./data
  --epoch EPOCH         Epoch for trainning. Default: 10
  --wvclass WVCLASS     Wordvector class, None for using Glove pretrained
                        wordvec. Default: None
  --wvpath WVPATH       Path for pretrained wordvector. Default: wordvec
  --out_dir OUT_DIR     Output directory for test output. Default: ./output
  --log_dir LOG_DIR     Log directory for tensorboard. Default: ./tensorboard
  --model_dir MODEL_DIR
                        Checkpoints directory for model. Default: ./model
  --cache_dir CACHE_DIR
                        Checkpoints directory for cache. Default: ./cache
  --cpu                 Use cpu.
  --debug               Enter debug mode (using ptvsd).
  --cache               Use cache for speeding up load data and wordvec. (It
                        may cause problems when you switch dataset.)

For hyperparameter settings, please refer to run.py.

TensorBoard Example

Execute tensorboard --logdir=./tensorboard, you will see the plot in tensorboard pages:

Following plot are shown in this model:

• loss: reconstruction loss + kl loss.

  

• perplexity: reconstruction perplexity.

  

• kl_loss: kl_weight * min(kld, min_kl=10).

  

• kld: kl divergence.

  

• kl_weight: weight to the kl loss in the loss function.

  

And text output:

Following text are shown in this model:

• args

Case Study of Model Results

Execute python run.py --mode test --restore last

The following output will be in ./output/[name]_test.txt:

perplexity:     7.292317
Fancy decorated bathroom with a toilet , sink , and shower .
A close-up of a plate with a sandwich in the middle .
Two giraffe stand together eating some leaves .
A street sign is displayed on the sidewalk .
A city street with cars and buses driving down it .
A painting with a clock tower in the middle of a city .
A man sitting at a table eating food .
A small airplane flying in the sky .
Two children in a field playing with a frisbee .
A train with graffiti on the side of it .
A woman walking in front of a bus .
A man is riding a wave with the ocean .
A television that is in the middle of a room .
Man in white shirt with tennis racket in hand .
The young girl is riding a horse through the water .
A man is holding the neck up as he holds his hands .
A glass vase filled with flowers in a room .
A man driving a motorcycle with a dog .
People are riding skis down a snowy slope .
A piece of cake is on a plate with a fork .
A persons hands holding up a cell phone in their hands .
A white and red plane is flying in the sky .
...

Performance

	Reconstruction Perplexity	KL divergence
MSCOCO	7.29	10.49

Author

Qi Zhu