Download Data

Download data for the experiments

At the root of the repository, run the following command to download the data files

git lfs install
git clone https://huggingface.co/datasets/saibo/GCD-data-v2
mv GCD-data-v2 data

Download the grammar objects

At the root of the repository, run the following command to download the compiled grammar files

git lfs install
git clone https://huggingface.co/datasets/saibo/GCD-grammar-v2 assets/pgf

Unzip the compressed grammar files

cd assets/pgf
# unzip and remove the zip files
unzip ED.zip && rm ED.zip
unzip CP.zip && rm CP.zip
unzip IE.zip && rm IE.zip

Get models

Create an environment variable HF_MODELS_DIR that points to the directory where you store the models.

For example, we create a directory ~/models and save the model of llama-7B by running the following command:

mkdir ~/models
# Make sure you have git-lfs installed (https://git-lfs.com)
git lfs install
git clone https://huggingface.co/saibo/llama-7B ~/models/llama-7B

Then, we set the environment variable HF_MODELS_DIR to ~/models by running the following command:

export HF_MODELS_DIR=~/models

The models such as LLAMA-7B need to be in HuggingFace format.

We don't provide other model weights as they are too large and may have licensing issues.