Thomas J. Fan

API wise, I'll be okay with flipping a coin during training time and adding a flag to trigger it so we stay backward compatible. > If no missing-values are encountered...

If there is a setup.py + src directory, then I assume there is a package to be installed with the src layout. There are two ways to make the `pyflyte...

> It seems like the mistral and phi3 configs were missed entirely? Any reason for that? I had a quick script to automate adding `max_seq_len` and missed mistral & phi3...

`get_params` only shows the parameters in `__init__` which does not contain the fitted attributes in an estimator. I see a possible way forward to resolve this issue: 1. Ensure all...

I opened https://github.com/scikit-learn/scikit-learn/issues/29893 to discuss the API implications of a `FrozenEstimator` on our API.

> And right now we explain "frozen" by using the word frozen a lot ♻️. And I struggle to come up with a better suggestion. If we are using the...

I am okay with making it public with the name `binary_classification_curve`.

Historically, we have added parameters such SVC's `break_ties` to help break ties and remain backward compatible. In this case, we can introduce a `break_ties` parameter to KNeighhbors with options: 1....

From triaging meeting: There are performance concerns with adding a tie breaking mechanism. For now, I am marking the issue as low priority and we may revisit in the future.

We recently added a way to configure `FuncTransformer` to get feature names: https://github.com/scikit-learn/scikit-learn/pull/21569