Currently running architecture, design, and coding test against the Qwen3-Coder-30B-A3B-Instruct running on local. Initially my PC crashed as I was running the Firestorm Viewer at the same time with the AI model and memory was consumed fast.
Next run was without Firestorm Viewer running. The PC didn’t crash this time but llama.cpp throws an error regarding GPU offloading. Some tweaks on execution parameters on llama.cpp with Vulkan support was successful.
Right now I have to asses how much complexty does this AI model can handle when it comes to creating softwares.
This site will hold short thoughts, daily observations, technical discoveries, project progress, and records that do not require a full article.