Want to help improve the What-If Tool?

[James Wexler]

Now that the What-If Tool codebase has moved out of the TensorBoard repository, and we're looking towards future improvements for the next major release, we want to open-up development to anyone interested in helping out.

Our development guide is up to date with instructions on how to build and run the tool in different environments (in notebooks, TensorBoard, or the web demos).

We've updated our issues list to tag issues that are good candidates for outside contributions with the following tags: "bug" (for bugs), "enhancement" (for new functionality), and/or "good first issue" (for bugs or new functionality that would be a good issue for someone new to the project to tackle).

If you are interested in taking any of those issues, just comment in the issue on github and we can start a conversation with more details about the work for that issue.

Here is a glimpse at the titles of those issues as of now:

- Automatic slice detection
- Create a feedback module
- Error message handling when remote model fails
- Multi-point editing
- Handle large images
- Style the global attributions table
- Add rank correlation when comparing two models
- Add sort by variation documentation
- Calculate/display AUC for PR curve
- Jupyter Lab 2.x support 
- Clarify connection between fairness optimization choices on left and data on right

- Data split for binning - Datapoint editor vs. Performance & Fairness

Also, feel free to create issues for other enhancements you are interested in, and we can discuss if it's feasible and possibly how it could be implemented.

Thanks for your interest in WIT and hope to work on some issues with some of you soon!