Torsten Wörtwein

Thank you! Looking forward to a faster `FaceLandmarker`! Not sure whether it is related: the speed up from switching from CPU to GPU for the `FaceLandmarker` is rather small (~15%...

Thank you @kuaashish for looking into the issue! Can you please specify what type of additional logs you are looking for? The above is a minimal reproducing example, demonstrating that...

> This fixes it, PR welcome? Yes - probably should use a more specific TypeVar, maybe `DataFrameT = TypeVar("DataFrameT", bound=DataFrame)`.

I opened https://github.com/pandas-dev/pandas/pull/58339 for pandas. Feel free to open a similar PR here

I'm fine accepting PRs that use it! I'm not sure whether that will create more maintenance burden on our side: - we might need to create/split overloads to pin-point the...

We should probably return `Self` more often: after double checking that pandas actually returns the sub-class, for example, `_consructor` returns `DataFrame` unless a sub-class overwrites it.

While they look very much private, they are documented: https://pandas.pydata.org/docs/reference/api/pandas.api.extensions.ExtensionArray._accumulate.html https://pandas.pydata.org/docs/reference/api/pandas.api.extensions.ExtensionArray._reduce.html and could therefore probably be added to pandas-stubs? @Dr-Irv

This seems to work now with mypy 1.4.1 and the latest pandas-stubs :)

You are right - I tested it without `--warn-unreachable`

using `str | None` sounds reasonable. Technically, any type seems to be accepted at runtime: ```py >>> df.index.names = [None] >>> df.index.names FrozenList([None]) >>> df.index.names = [1] >>> df.index.names FrozenList([1])...