Error on custom dataset

[israrbacha]

I tried with celebAmask-HQ dataset (MaskGAN) with 19 lebel_nc and got a strange error: RuntimeError: cuDNN error: CUDNN_STATUS_NOT_INITIALIZED terminate called after throwing an instance of 'c10::Error' what(): CUDA error: device-side assert triggered

after searching for a while i found that someone posted the same kind of error on "pix2pixHD" model but i am yet unable to solve it. kindly help if anyone knows. BTW Facades dataset work fine so there is no problem with CudNN or CUDA i think.

[TomatoBoy90]

I also got this erro,have you resolve this isssues?

[JanOdenthal]

For me the problem was, that I used RGBs as Input and not one-hot-coded labels. Using the "--no_one_hot" option solved the issue for me.

[faziletgokbudak]

For me the problem was, that I used RGBs as Input and not one-hot-coded labels. Using the "--no_one_hot" option solved the issue for me.

Does that mean your inputs are generic RGB images? What other options have you modified? I try to run the network for RGB or gray scale input images and don't know what to do with "--label_nc" option as, I assume, it defines the number of channels of input images.