Machine Learning Workspace(Windows 10)

written by dongheeh, hyejinp

1. Machine Learning Workspace Introduction

“All-in-one web-based development environment for machine learning”
This tool provides a docker image with an already built machine learning workspace. One of the advantages of this is web-based. Linux desktop GUI can access through a web browser. Using ml-workspace, it gives convenience to machine learning model development.

 Here is the ml-workshop github: https://github.com/ml-tooling/ml-workspace 
 

There are lots of ultimate tools for developers. Among them, we ran Jupyter Notebook which is web-based IDEs for data processing and Netdata for monitoring hardware status. We also verified by using pre-installed machine learning libraries such as pytorch, sklearn, pandas in the process.
 We are working on a movie recommendation model system project, so we tried to preprocess this model and data using this tool.

GitHub - ml-tooling/ml-workspace: 🛠 All-in-one web-based IDE specialized for machine learning and data science.

 

2. How to install

First, install the WSL2

1. In Windows Powershell

Executes a shell command in Windows Powershell.

Enable-WindowsOptionalFeature -Online -FeatureName Microsoft-Windows-Subsystem-Linux

2. In Microsoft Store

Install the Ubuntu 20.04.4 LTS

3. Reboot your computer

Second, install and setup the docker

1. here the download link and click the Docker Desktop for Windows

https://docs.docker.com/get-docker/

 

Get Docker

2. Reboot your computer

 

What is docker?

Docker is an open source project that runs and manages Linux applications as containers using process isolation technologies. Here's a quote from the Docker web page:

 

A Docker container wraps some kind of software in a complete filesystem that contains everything needed to run the software. This includes code, runtime, system tools, system libraries, anything that is installed on the server. This guarantees that it will always run the same regardless of the environment in which it is running.

https://en.wikipedia.org/wiki/Docker_(software)

Docker (software) - Wikipedia

 

Third, Getting Started Machine Learning Workspace 

 

1. Executes a shell command in cmd.

docker run -p 8080:8080 mltooling/ml-workspace:0.13.2

and then it will download many files.

 you will see this cmd state(don't close the cmd)

2. Open new explore or chrome etc.. and go this link (http://localhost:8080)

and you will see Web-based Jupyter Notebook 

3. Desktop GUI (Optional)

Open Tool -> VNC Click!

 

As the language setting of my browser is Korean, it is indicated as a "연결" button in the image below.
Click this button.

and then you will see below image

Enjoy!!

 

Apply to My movie recommendation system

1. Use Jupyter Notebook to data preprocessing

What is Jupyter Notebook?

Jupyter notebook is the one of web-based IDEs that allows you to run Python step by step. It can create and share documents containing live code, equations, visualizations, and narrative text and used for data cleaning and transformation, numerical simulation, statistical modeling, data visualization, machine learning, and more. In our project, data preprocessing is required to collect user responses. We can check this conveniently in Jupyter Notebook.

 

https://en.wikipedia.org/wiki/Project_Jupyter

Project Jupyter - Wikipedia

 

First.  Write code for data preprocess and Data upload

Second. Execute code on jupyter

2. Check my hardware state by using Netdata(optional)

What is Netdata?

Netdata is a real-time monitoring program to Linux, and the amount of hardware usage such as RAM of the computer currently in use is arranged in a table. We can easily check the figures in real-time every moment at a glance. After running the model, it allowed us to check our hardware status.

https://en.wikipedia.org/wiki/Netdata

 

Comment

It comes with the basic modules necessary to carry out a machine learning project, which is very convenient when starting a project. In addition, since it operates based on Docker, projects can be carried out regardless of a specific environment, so even novice programmers can easily build a machine learning environment. It helps.

However, from the point of view of an experienced programmer, it is not necessary. It accommodates more modules than expected, but because it is a workspace, memory can be wasted.
In addition, if a real-time streaming service is required to be applied to our movie recommendation project, it is better to run it in the host rather than serving it in the ML workspace.