Stability is key. If you want adoption for your compute hardware, it needs to provide reliable model coverage and performance. If you want to bring a novel edge AI product to market, you need to know if and how fast the application runs on your target device. Roofline’s performance dashboard provides the transparency metrics you need across all relevant models and devices.
With nightly updates and continuous model additions, you always work with the latest results.
For a given model, we track latency, throughput (e.g., tokens per second), and memory footprint as key metrics. For a given hardware, we track model coverage for an extensive model zoo as well as the metrics mentioned above.
The performance dashboard is updated on a daily basis. We have built extensive CI/CD infrastructure that performs nightly benchmarking runs for 100s of models across devices. The data is then fed into the performance dashboard.
Just reach out through our Contact Us form. We will get in touch immediately.
