papermill-mlflow

Simple data science experimentation with jupyter, papermill, and mlflow

Associated blog post: A simpler experimentation workflow with Jupyter, Papermill, and MLflow

Quick-start

• Clone this repo

git clone [email protected]:eugeneyan/papermill-mlflow.git

• Set up virtualenv

cd papermill-mlflow

# Create virtualenv based on requirements.txt
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Install kernelspec for Jupyter notebooks (the name argument must be identical)
python -m ipykernel install --user --name=papermill-mlflow

• Start Jupyter notebook

cd notebooks
jupyter notebook

• Run the cells in runner.ipynb

• Start MLflow (in another terminal)

# Open another terminal

# Activate the virtualenv
cd papermill-mlflow
source venv/bin/activate

# Start the mlflow server
cd notebooks
mlflow server

• Access the MLflow UI opening this in a browser: http://127.0.0.1:5000
  • Navigate to "indices" in the experiment tab if necessary