Closed alexbarghi-nv closed 2 months ago
For the multi-GPU examples, we set RAPIDS_NO_INITIALIZE to prevent RAPIDS from creating a CUDA context. However, this had no effect because there were some RAPIDS imports before that. This PR resolves that problem by moving those imports elsewhere.
RAPIDS_NO_INITIALIZE
For the multi-GPU examples, we set
RAPIDS_NO_INITIALIZEto prevent RAPIDS from creating a CUDA context. However, this had no effect because there were some RAPIDS imports before that. This PR resolves that problem by moving those imports elsewhere.