AI requires a lot of power...
In the midst of the modern world’s turmoil here I’m comparing tech specs of different cards suitable for AI tasks (Deep Learning, Object Detection and LLMs). They are all incredibly expensive though.
What is this trendy AI-assisted coding?
Vibe coding is an AI-driven programming approach where developers describe desired functionality in natural language, allowing AI tools to generate code automatically.
Whole set of MM* tools is on EOL...
I’ve used MMDetection (mmengine, mdet, mmcv) quite a bit, And now looks like it’s out of the game. It’s a pity. I liked it’s model zoo.
Comparing two deepseek-r1 models to two base ones
DeepSeek’s first-generation of reasoning models with comparable performance to OpenAI-o1, including six dense models distilled from DeepSeek-R1 based on Llama and Qwen.
Next round of LLM tests
Not long ago was released. Let’s catch up and test how Mistral Small performs comparing to other LLMs.
Awesome new AI model to produce image from text
Recently Black Forest Labs published a set of text-to-image AI models. These models are told have much higher output quality. Let’s try them out
So many models with billions of parameters..
Testing how Perplexica performs with various LLMs running on local Ollama: Llama3, Llama3.1, Hermes 3, Mistral Nemo, Mistral Large, Gemma 2, Qwen2, Phi 3 and Command-r of various quants and selecting The best LLM for Perplexica
Comparing two self-hosted AI search engines
Awesome food is the pleasure for your eyes too. But in this post we will compare two AI-based search systems, Farfalle and Perplexica.
Running copilot-style service locally? Easy!
That’s very exciting! Instead of calling copilot or perplexity.ai and telling all the world what you are after, you can now host similar service on your own PC or laptop!
Testing logical fallacy detection
Recently we have seen several new LLMs were released. Exciting times. Let’s test and see how they perform when detecting logical fallacies.
Not so many to choose from but still....
When I started experimenting with LLMs the UIs for them were in active development and now some of them are really good.
Requires some experimenting but
Still there are some common approaches how to write good prompts so LLM would not get confused trying to understand what you wand from it.
Labelling and training needs some glueing
When I trained object detector AI some time ago - LabelImg was a very helpful tool, but the export from Label Studio to COCO format wasn’t accepted by MMDetection framework..
8 llama3 (Meta+) and 5 phi3 (Microsoft) LLM versions
Testing how models with different number of parameters and quantization are behaving.
Ollama LLM model files take a lot of space
After installing ollama better to reconfigure ollama to store them in new place right away. So after we pull a new model, it doesn’t get downloaded to the old location.
