Replies: 1 comment 2 replies
-
Would you mind posting this question in discourse if it is a usage question or if it's a bug, post it as an issue? |
Beta Was this translation helpful? Give feedback.
2 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
I'm trying to calculate the svd of large matrix by using dask+rapidsai on gpus, refering to the following two links:
https://blog.dask.org/2020/05/13/large-svds
https://gist.github.com/quasiben/98ee254920837313946f621e103d41f4
I copied the codes from the referances, but it did not work, and returned
here is the python code (because I'm using one node, I have deleted some configurations about multi-node):
here is the error information:
It seems that the function is using numpy.ndarray (not cupy.ndarray) during the svd calculation. And the u,s,vh
are numpy.ndarray, which should be cupy.array.
By the way, the addition of two matrixes can be calculated on gpu:
Is anybody know how to fix this issue? Thanks.
Beta Was this translation helpful? Give feedback.
All reactions