Check out this repository

Patient health data needs to be provided in a csv file. It needs the following columns:

• date (eg. 15.12.17)
• time (eg. 13:59)
• bgValue
• cgmValue
• basalValue
• bolusValue
• mealValue
• glucoseAnnotation

Go to Folder cd data/csv

Add file called data.csv

Start with a file containing about 5 days of history and see how fast the computation is.

Go to Folder cd data/input (call it from root of project)

Add your profile.json and profile.pump.json files

Build the docker image from the provided Dockerfile

docker build -t t1d-pred:latest . (call from root of project)

You need to mount your local files and input data to the docker container

docker run --rm -v <absoulute path to root of project (should end with /t1d-prediction)>:/t1d t1d-pred:latest

Inside the docker container the python/main.py will be called.

Nodemon will watch inside the docker container for file changes (only .py) and will restart the main.py docker run --rm -v <absoulute path to root of project>:/t1d t1d-pred:latest nodemon --exec python python/main.py -e py -L

docker stop $(docker ps -a -q --filter ancestor=t1d-pred:latest --format="{{.ID}}")

To change code, open the python directory with your editor of choice. For rapid development start the docker container with autowatch. It will rerun the code everytime you change a python file.

After the first run, autotune has created all necessary files and run_autotune in python/main.py can be set to False for a faster runtime.

docker build -t t1d-pred-ml:latest -f Dockerfile-ML .

docker run --rm -v ~/code/t1d-prediction:/t1d t1d-pred-ml:latest

docker build -t t1d-pred-ml:latest -f Dockerfile-ML . && docker run --runtime=nvidia --rm -v ~/code/t1d-prediction:/t1d t1d-pred-ml:latest