Graph Attention Matrix Completion (Pytorch)

Includes [PyTorch Implementation] (https://github.com/tanimutomo/gcmc) of Graph Convolutional Matrix Completion (PyTorch and PyTorch Geometric)

Improves upon GCMC using graph separation and fusion layers.

Setup

• Setup a virtual environment with python 3.6 (or newer)
• Requires PyTorch 1.7 (or newer) and CUDA 10.2 (or newer)
• Install requirements (pip)

  pip install -r requirements.txt
  

Installation of Pytorch Geometric is very troublesome and may destroy your python environment.
So, we strongly recommend to use a virtual environment (e.g. pyenv, virtualenv, pipenv, etc.).
Please see Pytorch Geometirc official document for more details.

Train and Test

cd src
python train.py

• Configuration:
  The settings for train and test are in config.yml.

• Dataset:
  Training dataset is MovieLens-100k. The dataset is automatically downloaded in data/ by running src/train.py.