How to parallerize dual coordinet descent mehods on GPU using CUDA.jl?

You’re indexing a CuArray yourself, w̃[i], not sure what you expect.

The actual error tells you what the problem is and how to debug: If the returned value is of type Union{}, your Julia code probably throws an exception. You are also passing a CPU vector to a GPU kernel, that will never work. Neither will passing sparse vectors – CuSparseArray is an abstraction to work with the CUSPARSE library, and cannot be passed to kernels.

1 Like