/NFT-Recommender-NGCF

Primary LanguagePython

NFT-Recommender-NGCF

config

  • The user-defined parameters needed for the experimental setup are stored.
    • e.g., data path, epoch, batch size, metrics,
  • File name: fixed_config_{model_name}.config

dataset

  • collections
    • For each collection, there are .inter and .itememb files included.
    • The .inter file represents user-item interactions, and the .itememb file represents item embeddings.

hyper

  • Used for hyperparameter optimization.
  • hyperparameter search ranges are included.
  • File name: {model}.hyper

hyper_result

  • Contains the results of hyperparameter optimization.
  • The .best_params file contains the optimal hyperparameters that are saved after the hyperparameter optimization process.
  • The .result file contains the performance of all hyperparameter combinations.
  • File name: {model}_{dataset}.hyper

result

  • The results of the performance evaluation on the test set are saved in the result folder.
  • File name: {model}-{dataset}.csv

run.sh

  • Contains shell scripts that can be used to run the main file.

saved

  • The best model that shows the lowest valid metric during the model training process is saved.

Create_dataset.ipynb

  • Code to create the input data file in the format required by Recbole.

Create_config.ipynb

  • Code to create the configurations in the format required by Recbole.

main.py

  • The code to run experiments using our model.

Model.py

  • The code contains our model architecture.
  • The arguments are as follows:
    • dataset: the name of dataset (e.g., 'azuki', 'bayc', 'coolcats')
    • item_cut: the minimum number of interactions for an item to be used