MLGproject

article: https://medium.com/@nm8144/a7a40caeb959

Running

• If running from local, set env variable DATASET_PATH to path of your local dataset folder.
• If running from colab, just run and use Google GPU :)
• Running GAT: very important, to make sure concat=False in GATConv layer. Otherwise, it will not work. See here for more info.

Config

There are two configs:

• Test config: specify training configuration (epochs, batch size, etc.)
• Model config: specify model configuration (layers, activation functions, etc.)

It's best to re-run from Model definition cell to avoid errors and unexpected behaviour.

Todo

• DONE :)

Dataset

Is available at Kaggle