isabella232 / nlc2cmd

NLC2CMD

This repository contains our contribution to NLC2CMD Challenge

Requirements

Install python dependencies:

pip install -r requirements.txt

Clone clai repository in the folder nearby

git clone https://github.com/IBM/clai.git
git checkout -t remotes/origin/nlc2cmd

Download preprocessed manpage data cmd_options_help.csv

Training

To train the model(s) in the paper, run this command:

./train.sh <path to nl2bash-data.json> <path to manpage-data.json> <path to dev dir> <cmd_options_help.csv> <best clf model epoch> <best ctx model epoch>

for example

mkdir dev_dir
./train.sh nl2bash-data.json manpage-data.json dev_dir cmd_options_help.csv 4 6

Evaluation

To evaluate the model use tools from clai repository and submission_code folder.

Pre-trained Models

You can download pretrained models here: pretrained models

Results

Our model achieves the following performance on :

NLC2CMD Challenge

Model name	Accuracy	Energy (mW)
jb	0.499	828.9

Contributing

All content in this repository is licensed under the MIT license.