Flux, CUDA, Zygote : InvalidIRError: compiling kernel getindex_kernel(CUDA.CuKernelContext, CuDeviceArray

Haven’t figured out why ForwardDiff is being used here, but it I think it doesn’t work with FFT, e.g. yesterday: ForwardDiff and Zygote cannot automatically differentiate (AD) function from C^n to R that uses FFT

1 Like