News Recommendation Model using GNN.
Raw and preprocessed datasets can be accessed by this link from Google Drive: datasets.
You can use colab.ipynb
to run model on Google Colab or directly go to following link: colab.ipynb.
You need to sign in with Google account to use Colab.
If you don't want to download datasets, they can be can be mounted directly from the Google Drive within the Colab notebook. Add datasets from the 'Shared with me' folder into your Drive by clicking 'Add shortcut to Drive' to access it from the notebook. Steps on how to run the model is well documented in the notebook along with the code to setup necessary libraries. GPU hardware is available on Colab's free version.
Alternatively, main.py
can be run from local machine by downloading the datasets and installing libraries from requirements.txt
and PyG from its source:
pip3 install -r requirements.txt
pip3 install -q git+https://github.com/pyg-team/pytorch_geometric.git
With installed libraries, you can run main.py
from command line. For more options see --help
:
python3 main.py --data cit_pt --plm ptbert --epochs 50 --eval_steps 10 | tee logs_cit_pt_ptbert.txt