| layout | title | sidebar-title |
|---|---|---|
page |
WALA Ariadne |
WALA Ariadne |
WALA Ariadne uses static analaysis of Python for TensorFlow to give machine learning practitioners modern software development support. The video below gives a flavor of Ariadne, in this case integrated into JupyterLab:
- Integrated into common IDEs, such as JupyterLab
- Information loads asynchronously, displaying when available
- Information about tensor shapes displayed in code lenses and hovers
- Warning and error diagnostics using standard IDE mechanisms
- Quick fixes using standard IDE mechanisms
Right now, we support almost every environment and editor that utilizes the Language Server Protocol (LSP): Atom, Eclipse, Emacs, JupyterLab, NeoVim, PyCharm, Sublime, Spyder, Vim and VSCode are known to work to varying degrees.
Information on obtaining/installing Ariadne into your favorite IDE is available here, or click the logo of your favorite IDE below:
| 
| 
| 
| 
| 
|
| 
| 
| 
| 
| 
|
Through static analysis techniques (provided by WALA) such as Data Flow analysis, WALA Ariadne is able to provide type-like functionality that tracks tensors, provides backward slicing to track variables, and much more. We make this functionality available through the Language Server Protocol, meaning in most cases, it's as easy as a simple configuration to get any LSP-supported enviroment up and running.
The Ariadne framework is open source, hosted on GitHub, with projects for the analysis itself, the Language Server Protocol support and the underlying analysis framework.
We are currently looking for real life code examples to analyze. Leave it messy, leave it commented, we prefer to see the code in its natural state. Let us know what made it annoying or difficult to write, the bugs your encountered, and what kinds of support you'd like to see from a static analysis tool. Share your feedback here
- Slides - WALA Ariadne: Analysis for Machine Learning
- Paper - WALA Ariadne: Analysis for Machine Learning
