LEAP

Large Language Models as Interpolated and Extrapolated Event Predictors

Conda Environment

conda create -n leap python=3.9
conda activate leap
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2
pip install dgl==1.0.1+cu117 -f https://data.dgl.ai/wheels/cu117/repo.html
pip install -r requirements.txt

Event Interpolation

Detailed comments describing implementation steps can be found in each shell executable (.sh) file.

To run interpolated object prediction as a ranking task:

Please unzip "data.zip" file first.

cd LEAP_OP1
sh Train_LEAP_OP1.sh

To run interpolated object prediction as a generative task:

cd LEAP_OP2
sh Train_LEAP_OP2.sh

Event Extrapolation

To run extrapolated multi-event forecasting task:

cd LEAP_MEF
sh Train_LEAP_MEF.sh

References

Inspired by SeCoGD, we build LEAP_OP1.
Inspired by Glean, we build LEAP_MEF.