Nick Becker

Perhaps this would be a good first issue for a new contributor?

This seems to work locally for me on dense matrices with the existing machinery (using #5140). As a note, for sparse matrices we need to explicitly coerce the input data...

Are you running a multi-GPU workload? RAPIDS [pip packages don't yet support multi-GPU](https://rapids.ai/pip.html) (though conda packages / docker containers do).

[The dask-cuda documentation may be useful here](https://docs.rapids.ai/api/dask-cuda/nightly/examples/best-practices.html). But, I'd recommend using conda or Docker for now.

@sorenmacbeth , @stefanoviel could you share any details about how often you use this metric and how long it takes with CPU-based HDBSCAN (i.e., is this calculation on CPUs annoyingly...

This sounds like a loop carried dependency: ```python # Set the density sparseness of the cluster # to the sparsest value seen so far. DSC[label1] = max(length, DSC[label1]) ``` If...

@btepera @csadorf , this should now work via CPU fallback for folks using the [new zero code change required UX](https://developer.nvidia.com/blog/nvidia-cuml-brings-zero-code-change-acceleration-to-scikit-learn), right?

Game changing performance improvement :)

Thanks for surfacing this bug in our input/output data structure conversion utilities. While we work on a fix, you can hopefully work around this issue by calling `.values` on each...

Thanks for surfacing this issue. Could you share a [minimal reproducible example](https://stackoverflow.com/help/minimal-reproducible-example) that includes the full error? This could be due the same underlying issue as https://github.com/rapidsai/cuml/issues/5160 @dantegd @quasiben