Manuel Villanueva

Hi @jrp2014 , I just ran a install from zero to test this out but couldn't replicate it, if you could please share more detail on the install process you...

Currently working on a branch doing the CuFFT solution, I'm using our already implemented cuBLAS wrap currently being used for `mlx/backend/cuda/matmul.cpp` as reference on how are we handling CUDA libs...

Hi Leo, thanks for the idea! I'll be checking out the cuFFT `nvmath-python` Solution, and looking to use `mlx.core.from_dlpack` to have zero-copy and want to ensure stream alignment.

@leofang I'm starting work on this enhancement, would like to know if you have any highly recommended platform where I can test my changes with NVIDIA GPUs as I currently...