RuntimeError: cuDNN error: CUDNN_STATUS_BAD_PARAM

[RaySunWHUT]

hi, Mr. sefibk,

Thanks for your work, and if I use the "Blind data generator" to generate degradation images, the kernelGAN can estimate the correct kernel. But if the use a self-created disk kernel, and I run the code this a low-resolution image that downscales with the disk SR kernel. The code will occur error like this:

the disk kernel I used can be visualized like this:

and the kernel details is this:

def get_k():
    kk = [
        [0, 0, 0, 0, 0.0013339157770654528, 0.0036148217219170906, 0.0043595041514488267, 0.0036148217219170906,
         0.0013339157770654528, 0, 0, 0, 0],
        [0, 0, 0.00080003827014298045, 0.006060450060893929, 0.0088215133735283831, 0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088215133735283831, 0.006060450060893929,
         0.00080003827014298045,
         0, 0],
        [0, 0.00080003827014298045, 0.0076458238973070832, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0076458238973070832, 0.00080003827014298045, 0],
        [0, 0.006060450060893929, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.006060450060893929, 0],
        [0.0013339157770654528, 0.0088215133735283831, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088215133735283831, 0.0013339157770654528],
        [0.0036148217219170906, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0036148217219170906],
        [0.0043595041514488267, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0043595041514488267],
        [0.0036148217219170906, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0036148217219170906],
        [0.0013339157770654528, 0.0088215133735283831, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.0088215133735283831, 0.0013339157770654528],
        [0, 0.006060450060893929, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0088419412828830753, 0.006060450060893929, 0],
        [0, 0.00080003827014298045, 0.0076458238973070832, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753, 0.0088419412828830753,
         0.0088419412828830753,
         0.0076458238973070832, 0.00080003827014298045, 0],
        [0, 0, 0.00080003827014298045, 0.006060450060893929, 0.0088215133735283831, 0.0088419412828830753,
         0.0088419412828830753, 0.0088419412828830753, 0.0088215133735283831, 0.006060450060893929,
         0.00080003827014298045,
         0, 0],
        [0, 0, 0, 0, 0.0013339157770654528, 0.0036148217219170906, 0.0043595041514488267, 0.0036148217219170906,
         0.0013339157770654528, 0, 0, 0, 0],
    ]
    return np.array(kk)

[RaySunWHUT]

and by the way, if you can solve the last issue I mentioned, I will be very grateful!

[sefibk]

I think it is related to the CUDNN issue.
I can't think of a reason why the type of kernel should affect the run.
In fact - the algorithm is unaware of the kernel itself...