Vyas Ramasubramani
Vyas Ramasubramani
Reviewing this PR will be a bit unorthodox because of the experimental nature of this work, so we need to consider the desired outcomes on a few different timescales, the...
Yup I have those changes locally, let me push.
Thanks for a great discussion here y'all. I'm going to close this issue for now since it seems like all the original questions have been addressed, but feel free to...
Going to close as answered, but please feel free to reopen as needed. The UX here will continue to improve over time, but there are some things (like breakpoints) that...
/ok to test
Under the hood `value_counts` is basically a groupby-count. The reason for issue 1 is that cudf does not support the `observed` parameter (tracked in #15173). The underlying implementation can be...
Also yes Thomas is correct that we shouldn't expect pandas and cudf to produce identical ordering when `sort=False`. With cudf.pandas we may eventually try to change that using a compatibility...
Let's consider exactly what information we want to include before going too deep into the implementation here. @wence- I believe you originally made this request, WDYT?
When we resolve this, we can also address #16044