CSE 6250 BD4H - Final Project

Prerequisites

Dependency

Pytorch >=0.4
Python >=3.5

Dataset

MIMIC-III

Download MIMIC-III dataset and put DIAGNOSES_ICD.csv, PRESCRIPTIONS.csv, PROCEDURES_ICD.csv in ./data

Drug-Drug Interactions

Download Drug-Drug Interactions dataset and put drug-DDI.csv in ./data/

On-board the Project in Google Colab

To import the project in Google Colab, we recommend you to read below articles.

Run the Code

Modify Project Path

In the code, there are certain specified paths that currently access to our dedicated Google Drive directory.

You are able to customize the project path in your own Google Drive directory.

<CSE6250_Team_G3_final_submission.ipynb>

drive.mount('/content/gdrive')

sys.path.append("/content/gdrive/.../GAMENet/colab_env/lib/python3.10/site-packages")

path = "/content/gdrive/.../GAMENet/code_"
os.chdir(path)

End-to-End ML Pipeline

Open './CSE6250_Team_G3_final_submission.ipynb'
Runtime > Change runtime type > T4 GPU (or any available GPU)
Runtime > Run all

Training

Open './CSE6250_Team_G3_final_submission.ipynb'
Runtime > Change runtime type > T4 GPU (or any available GPU)
Run the End-to-End ML Pipeline to warm-up
In the section 'Running the Model - Train', adjust 'epoch_' to the target epoch (by default, a single epoch)
Run the cell
Model checkpoints and a final model will be saved in './code_/saved...'

Testing

Open './CSE6250_Team_G3_final_submission.ipynb'
Runtime > Change runtime type > T4 GPU (or any available GPU)
Run the End-to-End ML Pipeline to warm-up
In the section 'Running the Model - Test', modify 'resume_path' to the target model (by default, a pre-trained model will be loaded)
Run the cell
Evaluation will be executed.

taehyeok-jang/CSE6250-final-project