MNT Fix bug: FloatTransformer output dimension was set incorrectly.

[SeanMcCarren]

Description

Thanks to this elaborately typed out issue we found a bug in Adversarial Debiasing. Considering how big, highly parametrized, and not fully covered this implementation is, I'm surprised this is the first bug :)

The bug is simple. The FloatTransformer class's intention is to cast different types of input to a 2D array. But, in a specific case where no transformer is provided as input, and the output is not bi/multinomial, we accidentally set the output shape as the input shape.

Tests

• [ ] no new tests required
• [ ] new tests added
• [x] existing tests adjusted

Documentation

• [x] no documentation changes needed
• [ ] user guide added or updated
• [ ] API docs added or updated
• [ ] example notebook added or updated

[hildeweerts]

Many thanks, @SeanMcCarren! This looks good to me - but it seems the tests are failing unrelated to your PR (see #1334). After we fix that and the PR passes all checks I think it should be ready for merging.

[joelthe1]

Please consider merging and releasing this fix. Thanks!

[romanlutz]

Please consider merging and releasing this fix. Thanks!

@fairlearn/fairlearn-maintainers

[SeanMcCarren]

I am surprised this isn't merged yet, I am completely out of the loop of this. I can confirm that the issue raised here by parasurama will also be fixed by this PR.