Strata MoE Inference Engine Runs a 125B Qwen Model on a Single Gaming GPU | CodeYourCraft