Creating a Docker Environment for Jupyter Notebook-Based Machine Learning Projects - local on-premise setup guide

Aug 3, 2026

Upload by Noah

This project proposes configuration of a locally optimized development environment for Artificial Intelligence projects by combining containers with Jupyter notebooks. While Google Colab provides fast cloud access, subscription costs and resource limits can restrict certain workloads. The guide advocates building an on-premise, locally executable alternative that preserves Colab-style usability while enabling full hardware customization. Hardware selection, dependency installation, and environment setup steps are presented to help users gain control, flexibility, and GPU choices.

LanguageEnglish

CategoryTechnology

FormatPDF

Pages30

Reading Time1 h 16 min

AccessFree to download

Introduction

Choosing the Right Hardware for Machine Learning Environments

Nvidia GeForce RTX 4090: A low budget solution

Popular tools for ML Development

Jupyter Notebooks

Google Colab

Setting Up a Local Environment

Docker configuration

Link Google Drive as volume

Automatize the process

Conclusions

Bibliography

| Number | 296 | | --- | --- | | Publication Year | 2024 | | Acceptance in OA@INAF | 2024-03-21T14:57:53Z | | Title | Creating a Docker Environment for Jupyter Notebook-Based Machine Learning Projects | | Authors | CABRAS, Alessandro | | Affiliation of first author | O.A. Cagliari | | Handle | [http://hdl.handle.net/20.500.12386/35013](http://hdl.handle.net/20.500.12386/35013) ;
[https://doi.org/10.20371/INAF/TechRep/296](https://doi.org/10.20371/INAF/TechRep/296) | Creating a Docker Environment for Jupyter Notebook-Based Machine Learning Projects Authors ALESSANDRO CABRAS 1 1 INAF-Osservatorio Astronomico di Cagliari Reviewers ANTONIO PODDIGHE ANTONIETTA ANGELA RITA FARA Contents Abstract ii List of Acronyms iv List of Figures vi 1 Introduction 1 1.1 Choosing the Right Hardware for Machine Learning Environments ......... 2 1.2 Nvidia GeForce RTX 4090: A low budget solution .................. 3 2 Popular tools for ML Development 7 2.0.1 Jupyter Notebooks ................................ 7 2.0.2 Google Colab ................................... 8 3 Setting Up a Local Environment 11 3.1 Docker configuration ................................... 11 3.2 Link Google Drive as volume .............................. 13 3.3 Automatize the process ................................. 14 Conclusions 16 Bibliography 19 Abstract This project proposes the configuration of a locally optimized development environment for Artificial Intelligence projects, leveraging containers and Jupyter notebooks. While services like Google Colab offer quick and convenient access to pre-configured cloud resources for machine learning, subscription costs and resource limitations may be restrictive for some projects. To overcome these challenges, the creation of a locally executable environment similar to Colab is suggested, but deployable on-premise on a local server. This approach allows for full hardware customization, including GPU selection, and eliminates subscription cost constraints. In the following chapters, the necessary steps to configure this local environment will be outlined, starting from hardware selection and proceeding with the installation of required dependencies and environment setup. By following these guidelines, users will be able to establish a local machine learning development environment that provides greater control and flexibility, while retaining the convenience and familiarity associated with Google Colab. List of Acronyms AI Artificial Intelligence GPU Graphics Processing Unit ECC Error-Correcting Code HPC High Power Computing CUDA Compute Unified Device Architecture PCIe Peripheral Component Interconnect express OAC Osservatorio Astronomico di Cagliari TPU Tensor Processing Units List of Figures 1.1 Two GPU NVIDIA RTX 4090, installed in a Supermicro server at the Osservatorio Astronomico di Cagliari (OAC) .............................. 4 1.2 ZOTAC GAMING GeForce RTX 4090 Trinity graphics card [1] ........... 4 1.3 One of the graphs extracted from the web comparing Graphics Processing Unit (GPU) performance in the market relative to their price. [2] ............. 5 2.1 Descriptive diagram of Google Colab, from user authentication to cloud resource allocation.......................................... 8

FAQs

Why build a local Docker environment instead of using Google Colab?

Google Colab can be limited by subscription costs and constrained resources. A local on-premise setup removes these constraints while keeping a familiar workflow for Jupyter-based development.

What steps are required to set up the local environment?

The document outlines selecting appropriate hardware, configuring Docker, linking Google Drive as a volume, and automating the setup process, followed by installing the needed dependencies.

How does hardware customization work in the proposed approach?

The local, container-based environment can be deployed on a local server, allowing users to select and configure hardware choices such as GPU options and related dependencies.

Aug 3, 2026

Upload by Noah