Setting Up Ollama with ROCm and Docker on an AMD Ryzen AI Max+ 395 System

·2 views

Running large language models on the integrated AMD Radeon 8060S requires the operating system, ROCm runtime, container setup, and GPU permissions to function together, since a container can start while inference silently falls back to the CPU. The author documents a setup on an EVO-X2 machine with a Ryzen AI Max+ 395 using Ubuntu, ROCm, Docker, and Open WebUI, along with methods for verifying GPU detection. The notes are drawn from late 2025 and were not fully reproducible because earlier container tags were floating and package versions were not fully recorded. The guide covers Docker and group permission steps, plus a ROCm 7.1.1 installer path that is presented as historical rather than a recommendation to downgrade. Readers are advised to consult AMD's compatibility matrix for supported Ubuntu and kernel combinations before installing.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

LLMOps: Why Shipping AI to Production Is an Operations Problem

A chatbot on Air Canada's website wrongly told a customer he could claim a bereavement discount after travel, contradicting the airline's own policy page. When the airline refused the refund, British Columbia's Civil Resolution Tribunal ruled in February 2024 that the company was responsible for its chatbot's statements and ordered it to pay C$812.02. The article argues that a deployed model is effectively a vendor dependency whose outputs carry the company's liability, so production readiness requires operating it, not just building it. It identifies four metrics to track: p95 latency, error rate, cost per 1,000 requests, and eval pass rate. It presents these as the concrete measures that distinguish a live product from a demo, and frames the piece as the final part of an AI engineering series.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Builds C++ Monte Carlo Stock Price Simulator Using CMake

A developer has built a stock price simulator written in C++ with CMake as its build system. The project's main purpose is to help the creator gain hands-on experience with both C++ and CMake. The tool forecasts possible future stock prices by starting from the current price and applying random shocks drawn from a normal distribution. The developer is currently adding support for historical stock market data and has shared the code on GitHub, requesting feedback from readers.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Guide Shows How to Export Instagram Post Comments to CSV Using an API

A developer who builds the SocialHolmes social data API explains why collecting comments from public Instagram posts is difficult, citing login walls, rate limits, frequently changing internal queries, and paginated loading. The article reviews three approaches: the official Instagram Graph API, which only covers accounts connected to the developer's app, do-it-yourself scraping, and paid scraping services. It then walks through a curl workflow that starts a scraping run, polls its status, and reads results in pages of up to 100 comments. A Python script is provided that automates the process and saves the output to a CSV file.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Shares Free, Local AI Setup Using Ollama as Alternative to Cloud APIs

A DEV Community post argues that cloud AI services charge per token and send user data to remote servers. The author presents Ollama as a way to run language models such as llama3.2 locally on a laptop at no cost. The piece includes setup commands and a short Python example that queries the local model through a simple HTTP request. It also lists several models suited to different tasks, including code, lightweight work, and complex reasoning. The author says they built 45 open-source tools with this approach.

0 comments Read more at DEV Community