Preparation

Create virtual environment

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

Download dataset

wget -P data/ -nc https://raw.githubusercontent.com/uiuc-cse/data-fa14/gh-pages/data/iris.csv

Checkout to experiments branch

git checkout -b experiments

Run demo

Set Google credentials

export GOOGLE_APPLICATION_CREDENTIALS=/path/to/your/credentials/json

Run demo notebook

jupyter-notebook demo[-remote].ipynb

Notebooks structure

demo

Part I. Usage mlpanel from notebook

Demonstrates how to communicate with tracking server running by mlpanel directly from Jupyter notebook.

In this part first function which used in ML-workflow are defined. The main function is run_experiment() - creates new run on tracking server.

Part one has screenshots of UI usage: view projects, experiments, runs, register models and create deployments.

Part II. Usage mlpanel from code

There is ML project example - IrisProject. It's MLflow project ready to use.
In Part II IrisProject logs parameters, metrics, artifacts and models to tracking server running also by mlpanel.

---

Both parts shows that it's not difference where from to communicate with mlpanel - main you must specify MLflow tracking uri for you project and run it.

demo-remote

Part I and Part II

Like in demo

Part III

About how to predict data