/program-synthesis-rl-tensorflow

in progress

Primary LanguagePython

Leveraging Grammar and Reinforcement Learning for Neural Program Synthesis

TensorFlow implementation of Leveraging Grammar and Reinforcement Learning for Neural Program Synthesis.

introduction

Requirements

Usage

Prepare with:

$ pip install -r requirements.txt

To generate datasets:

$ python dataset.py --data_dir=data --max_depth=5

To train a model:

$ python main.py
$ tensorboard --logdir=logs --host=0.0.0.0

Results

Currently, only maximum likelihood optimization is implemented. Expected correctness and RL is in progress.

(in progress)

References

Author

Taehoon Kim / @carpedm20