You can install the RAPIDS "core" as an extra python dependency, since NVIDIA provides
cuml built on CUDA 9.2. Create a
floyd_requirements.txt file and add this line:
or you can run
pip install cuml-cuda92 from a Workspace terminal. For the other extensions(such as
dask-cudf), you can download and build them from their repo (e.g.
pip install git+https://github.com/rapidsai/dask-cudf).
We will definitively add RAPIDS to our environments in the next updates.
Note: Make sure to run in a CUDA 9.2 environments such as
Hope that helps.