DistilBERT4Rec

To reproduce this work, please follow the following steps.

Step 1

Clone the repository

Step 2

This work has been done on the Movielens ML-1m and ML-20m dataset. To download the dataset, just execute the data_download_script.sh script. It can be executed using the following command -

sh data_download_script.sh

The script now only has the ML-20m download instructions. Similar instructions can be added for ML-1m. Alternatively, these datasets can be downloaded manually and used as well.

Step 3

Create a virtual environment using python -m venv <venv_name>

Activate the virtual environent using source <venv_name>/bin/activate

Install all the dependencies in the virtual environment using pip install -r requirements.txt

Step 4

To train the model run the following command at the project's root directory -

python run.py train <data_csv_path>

To test the model run the following command at the project's root directory -

python run.py test <data_csv_path> <trained_model_path>

The trained model path argument is necessary because this code is structured to evaluate on a trained pytorch model.