Bradley Dice

I noted this offline -- let's try to write "NDS-H" anywhere it appears in text, rather than "NDSH."

@Matt711 The PR description says "do not merge" but there is no "DO NOT MERGE" label. Can you make this consistent? Also for team knowledge, the "Description" section of the...

@jrhemstad @vyasr We have three patches, and I see two fixes. I see how the CCCL offset-dispatch changes will fix [`thrust_disable_64bit_dispatching.diff`](https://github.com/rapidsai/cudf/blob/branch-24.12/cpp/cmake/thirdparty/patches/thrust_disable_64bit_dispatching.diff) and `ChainedPolicy` changes will fix [`thrust_faster_scan_compile_times.diff`](https://github.com/rapidsai/cudf/blob/branch-24.12/cpp/cmake/thirdparty/patches/thrust_faster_scan_compile_times.diff). I don't see...

Yes, I included the relevant parts in #17233. Thanks for the help! Closing.

This probably needs updated since pylibcudf has been moved.

Thanks @vyasr - I had the same concern. Let's try to reduce our Arrow dependencies first, and we can upgrade this later in the 24.10 release cycle if still needed.

@raybellwaves Following up on this. Did you install `dask-cudf` before building the documentation? The docs builds require both `cudf` and `dask-cudf` to be installed so that the notebooks can be...

@galipremsagar Could you take this on?

Here's a better reproducer. This works on systems with V100 GPUs or others with less memory than an A100: ```python import rmm import cudf import cupy rmm.reinitialize(managed_memory=True) print("Creating series...") nrows...

This calls through a groupby-count.