Deep Dive into LLMs like ChatGPT with Andrej Karpathy
This is a general audience deep dive into the Large Language Model (LLM) AI technology that powers ChatGPT and related products. It is covers the full...
Video Tutorial
Summary
The video shows how to install and run the large language models Qwen‑3 235B (using Q3_K_XL) and GLM 4.5‑Air‑106B (using Q6_K_XL) on AMD Ryzen AI MAX “Strix Halo” hardware. It walks through kernel configuration, unified memory tuning, and provides benchmark results, with all scripts shared via a GitHub repo.
Highlights
- Step‑by‑step guide to enable the new AMD Zen 4C‑based Strix Halo AI acceleration on Linux
- Kernel build flags and driver settings required for Q6_K_XL / Q3_K_XL inference
- Unified memory (up to 128 GB) configuration and performance tuning tips
- Benchmark comparison of Qwen‑3 235B vs GLM 4.5‑Air‑106B on the same hardware
- All scripts, container images and benchmark logs posted publicly on GitHub
auto-generated
Context
Audience
AI engineers, Linux system administrators, and hobbyists who want to run state‑of‑the‑art large language models on AMD Ryzen AI MAX hardware
DomainMachine Learning / AI Inference
Formatvideo tutorial with accompanying code repository
Accessfree online (YouTube + open‑source GitHub repo)
Topics
Related
AMD Strix Halo developer documentationHugging Face Transformers inference guideFedora container tooling for AI workloadsLLM quantization techniques (Q3_K_XL, Q6_K_XL)HP Z2 G1a Mini workstation specifications