Important

If you'd like to run this code, you will have to gain access to the DiaTrend dataset at DiaTrend@Synapse.

How the code is structured:

The data folder would hold the original and cleaned datasets. The original dataset is a set of 54 .xlxs files. The cleaned dataset is a single subject's cleaned data that can be trained on by the model.

Note

If you want to run the code, you will have to unzip the DiaTrend dataset from DiaTrend@Synapse. Additionally, you'll have to exectue the clean_data.py script for your intended subject file(s).

The models folder holds different trained models. The file naming structure indicates the model architecture and model training specifics.

The src folder holds the model.py, train.py and clean_data.py scripts. model.py is the PyTorch model object. train.py is a one-off script w/ argument parsing for CLI training. clean_data.py is a one-off script w/ argument parsing for CLI data cleaning.

How to run the code:

Important

Assuming development environment in Linux, MacOS, or WSL.

Virtualenv

Note

In the diatrend root directory.

Important

Assuming a Python 3.12.8 version. Package conflicts may result if this is not the case.

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

TODO: create Dockerfile for containerization

Data Cleaning/Preprocessing

Note

In the diatrend/src directory

python clean_data.py --verbose --subject_id 4 --to_csv

Model Training/Testing

Note

In the diatrend/src directory

python train.py --epochs 10 --batch_size 32 --learning_rate 0.0001 --emb_dim 64 --num_heads 8 --ff_dim 128 --verbose --output_dir "../output"

Contributors:

• Nathan Panzer
• Daniel Arinze