• Posts
Introduction
Restructuring
Full stack deep learning
• Lab Modules
  • Lab 1
    • Pytorch and setup
• Lectures
Serverless ML
• Lab Modules
  • Lab 0
  • Lab 1
    • End to End
    • Refactoring
• Lectures

FSDL Lab 1: PyTorch

10/10/2022 Lab 1 - Deep Neural Networks in PyTorch #machinelearning #course #fsdl Follow along with the Lab 1 notebook. Or check out the course website for the Lab 1 slides. Setup Using a bootstrap python file, the Colab environment is setup to pull the FSDL github repo and set the path variable, lab directory, hot reloading of modules, and inline plotting. Downloading data We will be using the mnist dataset of handwritten digits.

October 10, 2022 Read

Serverless Machine Learning Module 0

05/10/2022 Intro We’re looking at lab 0 of the Serverless-ML course today. Make sure to check out their website and their Github repository and give them a star! A link to my fork of the repo is here if you wish to follow along with my notebooks! This course aims to allow a data scientist to deploy machine learning and predictive services online without the use of a server. This is useful in creating powerful business applications to show the predictive capabilities of a machine learning model quickly and effectively.

October 5, 2022 Read

End to end pipeline

05/10/2022 Lab 1 End to end pipeline #machinelearning #course #hopsworks #serverless #gradio We’re looking at lab 1 of the Serverless-ML course today. Make sure to check out their website and their Github repository and give them a star! A link to my fork of the repo is here if you wish to follow along with my notebooks! Introduction Following on from Module 0 labs we are now going to look into deploying an actual machine learning product to the web!

October 5, 2022 Read

Refactoring the pipeline

06/10/2022 Refactoring Our little web app in the previous article is nice but it’s quite rigid in structure, very reliant on the notebook, doesn’t allow for new data to be added, and can’t work standalone. Let’s take some steps to refactor this pipeline and deploy it in a better way. The new architecture for the pipeline we wish to implement will look like this ! Fig 1.

October 5, 2022 Read

Depth estimation project review

Introduction Today we’re going to take a look at a computer vision project on depth estimation and see how we can improve it. The goals of this investigation are twofold Highlight the good and bad practices used in the project Apply MLOps techniques to serve a production model in the cloud A few of the techniques to be included are the following: Unit testing, CI/CD, cloud serving, monitoring, and model experimentation and evaluation.

June 27, 2022 Read

Restructuring a project

Introduction Jumping right in, let’s start by creating a new folder on the desktop, a new conda environment, and activate that environment. mkdir ~/Desktop/DepthEstimation conda create --name depth_env conda activate depth_env I will be using the CookieCutter Data Science template to create a new project structure. This is done with the following commands: pip install cookiecutter cookiecutter -c v1 https://github.com/drivendata/cookiecutter-data-science After filling in the CookieCutter template, we can navigate to the newly created project and start working on it.

June 27, 2022 Read