The structure of the project is as follows:
gbpnet.models
: The implementation of the models used in the experiments.gbpnet.datamodules
: The data processing pipeline.run_test.py
: The main script for reproducing the results.
The verified versions of the dependencies can be found in requirements.txt
.
The datasets are automatically downloaded when the test script is called. The datasets will be stored in data/
directory.
We provide a demo model in the models
directory, which can be used to evaluate the results for CPD task.
python run_test.py ./models/cpd_model_sample.pt cpd
The following packages/libraries are adapted/communicated with in the codebase of GBPNet:
- Pytorch
- Pytorch Lightning
- Pytorch Geometric
- NumPy
- Catalyst
- Ingraham, et al, NeurIPS 2019
- Townshend et al, NeurIPS 2020
- Jing, et al, ICLR 2021
We thank their authors for providing the codebase.