Was able to make an LLM work on integrated Intel GPU (Core Ultra 7 155H) on Fedora. Got about 25% improvement compared to running the model on a CPU.

I'll share more in an upcoming blog post.

#ai #llm #intel #linux