Inference performance metrics across different models, hardware configurations, and serving parameters.
