Hello there 👋, This is a showcase of a project I’ve been working on for the past weeks. It’s a simple cross-platform desktop application that allows you to tag images using a machine learning model that I’ve trained. It is written in C# using Avalonia UI, and it uses the FastAI library for the machine learning part, the model is then converted to ONNX. The application is available for Windows, Linux and MacOS....

Hello there 👋, In this post, I will show you how to convert a FastAI model to ONNX and CoreML. Convert a FastAI model to Pytorch Before we can convert the model to ONNX format or CoreML, we need to convert it to Pytorch first. This is a simple process. Assuming that the model is trained, we need to call torch.save to export the model: from fastai.vision.all import * # learn is your FastAI Learner torch....

Hello everyone 👋, In this article, we’ll instrument a simple Node.js API with OpenTelemetry traces. OpenTelemetry is the new standard for distributed tracing and metrics. Trace data is used to monitor and troubleshoot complex distributed systems. At the end of this article, we’ll have a Node.js API that is instrumented with OpenTelemetry and we’ll be able to see the traces in Grafana Tempo: The API Let’s start by setting up our NodeJS project by installing the dependencies:...

Hello everyone! 👋 This is a follow-up post to the previous one, long story short I started taking the course from fast.ai to learn more about Deep Learning. I’ve built a simple Image Tagging model using the Resnet architecture (don’t worry about the paper I did not read it, GitHub Copilot suggested it while writing this blog post). FastAI is a high-level Python library that allows you to train complex ML models really fast and efficiently....

Hello everyone 👋, I’ve been playing with FastAI and deep learning for a week or so and I’ve decided to combine my passion for software with my passion for photography. I’m working on an app or maybe a library that allows you to generate tags for a photo. For an amateur photographer like me it’s quite useful as I don’t always have inspiration to write tags for my photos and sometimes the tags I write are inconsistent....

Hello everyone 👋, In this quick post I’ll show you how to deploy a FastAI model using FastAPI. It’s a follow-up to my previous post from here and it is partially based on this FastAI lecture: Exporting the Model Before we create the API, we need to export the model as a pickle file. Add learn.export("<path>/cat_or_dog.pkl") with the full absolute export path at the end of the script: # ... learn = vision_learner(dls, resnet34, metrics=error_rate) learn....

Hello everyone 👋, This is my first blog post about AI, it’s a quick post on how to enable GPU support for the FastAI library. I have avoided getting into AI for some time now, I thought that it would involve lots of complicated math, and I was not sure that it would be worth the effort, I did not get very good grates in my college math classes 😅....

Hello everyone 👋, Today I want to share with you a problem I’ve encountered along my journey. The problem is fairly simple: You have a finite number of water units. You have a finite number of buckets. You need to distribute the water units into the buckets. Each bucket has a maximum capacity. When a bucket is full, the water overflows into the next two buckets by splitting evenly. You need to print the number of the buckets that are full....

Introduction Hello everyone 👋, This is a quick post on how I migrated my blog from WordPress to Hugo. I will not go into details on how to install Hugo or how to use it, but I will share some of the things I learned along the way. I’ve also written a simple tool in Python to aid you in the migration process. You can find it here: https://dnutiu.github.io/jekyll-to-hugo/ Why not WordPress?...

The CAP theorem, also known as Brewer’s theorem, is a concept in distributed computing that states that it is impossible for a distributed system to simultaneously provide all three of the following guarantees: Consistency: All nodes in the system see the same data at the same time. Availability: Every request made to the system receives a response, without guarantee that it contains the most recent version of the data. Partition tolerance: The system continues to function despite arbitrary message loss or failure of part of the system....