Helper script for starting and viewing MLflow server from NERSC JupyterHub

start_mlflow.sh

#!/bin/bash

# Usage:
# - Go to https://jupyter.nersc.gov and start a server (e.g. login node)
# - Start a terminal from the launcher
# - Load your environment with mlflow installed and run this script:
#     conda activate <my environment>
#     ./start_mlflow.sh
# - Wait a few seconds for the server to start, then open the URL that is printed by the script.

port=${MLFLOW_PORT:-8123}

if [ ! -z $JUPYTERHUB_SERVICE_PREFIX ]; then
    jupyter_url="https://jupyter.nersc.gov${JUPYTERHUB_SERVICE_PREFIX}proxy/${port}/"
    echo "Once it starts, your MLflow server will be viewable at ${jupyter_url}"
fi

set -x
mlflow server -p $port