Self-Hosting

How PCIe Lanes Affect LLM Performance? Depending on the task. For training and multi-gpu inferrence - perdormance drop is significant.

The next step is to implement some CI/CD and setup Gitea Actions for Hugo website, to push website to AWS S3 automatically when master branch is updated.

In the Ollama models library there are models that able convert HTML content to Markdown, which is useful for content conversion tasks.

Will list here some AI-assisted coding tools and AI Coding Assistants and their nice sides.

I’ve got a theory to test - if utilising ALL cores on Intel CPU would raise the speed of LLMs? This is bugging me that new gemma3 27 bit model (gemma3:27b, 17GB on ollama) is not fitting into 16GB VRAM of my GPU, and partially running on CPU.

MinIO provides two standard UIs: (1) MinIO Console - That’s a web UI, and (2) MinIO Command - this is a commandline util mc.

MinIO is an open-source, high-performance, distributed object storage system designed to be fully compatible with the Amazon S3 API.

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.

Let’s see on github the list of best themes for Hugo static site generator, And check their star rating. You can go to official hugo themes list: https://themes.gohugo.io/

When the Ollama server receives two requests at the same time, its behavior depends on its configuration and available system resources.

Probably shoud call this platform engineering now. Here are some notes on how to install connect and remove the Portainer - docker images and containers manager UI.

Short comparison of the RabbitMQ on AWS EKS and AWS SQS

• features and costs.

Not every second Friday, but still… We have two mainstream approaches

Here’s a set of instructions on How to Configure to Run any Executable as a Service in Linux with Systemd.

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.