Efficient Constituency Tree based Encoding for Natural Language to Bash Translation

Setup

To create environment from tree.yml:

conda env create --file tree.yml

Training

To train the model in the paper, run:

python main.py --mode train

The default values of hyper-parameters are set to those mentioned in the paper. The full hyper-parameter list can be accessed with:

python train.py --help

Generate Bash Commands

To generate Bash commands for the invocations in the test set using the trained model, run:

python main.py --mode evaluate --checkpoint_path path_to_checkpoint_directory