Recent articles filed under Machine Learning

How to enhance your C# application with a large language model using Semantic Kernel

Created 5/29/2023 9:03:31 AM
Filed in Machine Learning

While chat applications have received much attention, there is so much more you can achieve by integrating language models into your projects. Imagine automating budgeting, generating engaging tweets, transforming documents, or creating compelling conference abstracts. Semantic Kernel offers a lightweight SDK that simplifies the process, allowing you to build language applications effortlessly. In this post, we'll explore how you can leverage Semantic Kernel to add powerful language skills, from installation to building skills and integrating them into your C# application. Unleash the potential of large language models and take your projects to new heights.

Fun ChatGPT prompts to improve your Python code

Created 4/23/2023 7:15:42 PM
Filed in Machine Learning

Let's explore fun prompts for improving your Python code using GPT-4. These prompts include writing docstrings, writing automated tests, checking your code for optimizations and bugs, and refactoring a method. While GPT-4 could be better, it's a significant improvement over doing these tasks by hand.

Why You Need a Sandbox Separate from Your ML Pipeline to Create Successful Experiments

Created 2/27/2023 7:08:58 PM
Filed in Machine Learning

Machine learning operations (MLOps) is a rapidly evolving field that is transforming the way businesses approach data science. One of the most critical aspects of MLOps is the ability to create and test experiments quickly and efficiently. While it may be tempting to use your production ML pipeline as a testing ground, it's essential to have a separate sandbox environment for several reasons.

Github Copilot: What you need to know to get started with AI-assisted programming

Created 2/16/2023 8:21:39 AM
Filed in Machine Learning

We’re starting with a pilot to test Github Copilot at Info Support. This is awesome news for us because we can now learn how much copilot is going to help us with our code. In the past few months, I’ve been testing Github Copilot on my own projects and learned a ton about things like use cases and plugins for various IDEs.

Did you know DVC doesn't handle large datasets? Neither did we and here's how we fixed it

Created 1/13/2023 9:37:46 AM
Filed in Machine Learning

We hit a nice roadblock yesterday while working on a dataset with 5 million samples. It took us 8 hours to upload the dataset and 11 hours to download and unpack it. We learned some valuable lessons there that I’ll discuss in this post so you don’t have to run into the same issues.

How to configure memory limits in WSL2

Created 1/5/2023 8:01:11 AM
Filed in Machine Learning

One of the downsides to building machine learning models is that you need quite a bit of power in your machine. I've just upgraded my home desktop that I use for side projects and experiments to 64GB of memory because it was clearly no longer up to the task. However, when I booted up my WSL linux distro, it couldn't get more than 32GB of memory. Time to figure out how to fix that! In this quick post I'm showing you how you can limit or increase the memory available to your WSL2 distributions. Let's go!

How to build scalable workflows with Prefect and Ray

Created 12/23/2022 7:38:36 AM
Filed in Machine Learning

In the past weeks we've been working on a new computer vision project at Aigency. As part of the project we have to process images using OpenCV image filters. This is a rather tedious process that takes a lot of time, so we made a nice workflow for it. In this blog post I'll show you how we used Prefect and Ray to build a scalable image processing workflow.

How to set up CUDA and PyTorch in Ubuntu 22.04 on WSL2

Created 11/2/2022 6:24:44 PM
Filed in Machine Learning

Working with machine-learning code in Windows can be quite a hassle. Often I run into problems with my tools not working properly on Windows. Recently, I even decided to convert my laptop to Ubuntu 22.04 for the time being. Which is really nice for data sciency stuff, but not great for my Powerpoint and Excel since LibreOffice sucks big time. Needless to say, I needed another solution. Especially for my deep learning setup. In this quick post I'll show you how I configured my Ubuntu 22.04 installation on Windows 11 to run CUDA and PyTorch.

Training machine-learning models with DVC

Created 10/17/2022 8:58:56 AM
Filed in Machine Learning

I've shown how to use DVC to manage data for your machine-learning project on my blog before. Today it's time to take things to the next level. Let's build a pipeline to train a machine-learning model with DVC.

Managing machine-learning datasets with DVC

Created 10/14/2022 9:59:21 AM
Filed in Machine Learning

DVC is a data versioning tool that is designed to make it easier to work with data for machine-learning projects. It integrates with git so that you can easily keep track of your data files and track changes to them over time. It also allows you to easily share data files between different projects and collaborators. In this post I'll show you how you can start using DVC in your ML project.

How to debug Airflow DAGs in VSCode

Created 9/10/2022 6:02:00 PM
Filed in Machine Learning

After last week's success setting up Apache Airflow I figured it would be nice to have a development setup so I can debug my own stuff. In this post, we'll look at how you can use VSCode to setup debugging for your Airflow DAGs.

How to deploy Airflow with Kubernetes and GIT

Created 9/8/2022 7:04:31 PM
Filed in Machine Learning

When I started working on MLOps a couple of years ago I ran into Azure Machine Learning first, because that's what the community around me was using. Turns out, it's not the only product out there. I already knew this, but I Ignored other products because we were happy with Azure ML at the time. This may sound like I'm no longer happy with Azure ML and want to move away. But that's not what this is about. I think you should always explore other technology options from time to time. That's why I'm trying out Airflow on Kubernetes.