Skip to content
code for Gulasch Programmier Nacht 2019 Workshop
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

GPN 19 - Talk on ML Workflow Tools

Modern ML workflow requires to run experiments fast at a large scale. In order to stay sane and keep an overview of what is going on, there are some tools out there.

We will demonstrate 2 tools in this workshop : ML flow and edflow. These tools focus on different parts of the development workflow : fast model iteration and monitoring (dashboard).

We will provide some basic code and instructions on how to add the tools to basic examples. We will cover a standard classification based problem to demonstrate a simple use-case.

Environment setup

  • create new virtual environment, for example using conda
conda create --name ml_talk python=3.6
source activate ml_talk
  • then install requirements
pip install -r mlflow_excersize/requirements.txt
pip install -r requirements_edflow.txt


Part 1 - EDFlow

Solutions for Edflow Part

edflow -t problem1_solution/train.yaml # train model
edflow -t problem1_solution/train.yaml -p project_folder # continue training model

edflow -t problem2_solution/train.yaml -e problem2_solution/validation.yaml # add validation

Part 2 - MLFlow

It consists only of adding some simple logging statements to mlflow_excersize/ and mlflow_excersize/ (compare mlflow_excersize/ and afterwards running mlflow ui --filestore mlflow_excersize/mlflow.

Part 3 - Project Structure

The template is available under:

You can’t perform that action at this time.