RuntimeError: cuDNN

[Limbicnation]

Hi I get the following error when running the code:

RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED when calling backward()

I would appreciate your help on how to resolve this.

Thank you!

Gero

[NNNNAI]

Sorry for the late respond ,did you follow the installation section in readme.md to build your environment? And what kind of GPU did you use.

[Limbicnation]

Hi there. No worries!
I did try to follow the installation instructions. However, I did get an error when building the environment with python 3.6.
I am using an RTX 3090 card.

PS: I am not in office at the moment, but I will try the installation instructions again in January.

Thanks for the reply!

[NNNNAI]

Oh, I suggest you to use CUDA >=11.1,because the defualt cuda in the installation instructions is 10.1 which is not suitable for 30 serial GPU

[Limbicnation]

Thanks for the information, I will try that!

[Limbicnation]

Hello,

I am still getting an error:

raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

Also I have installed CUDA 11.5

| NVIDIA-SMI 495.29.05 Driver Version: 495.29.05 CUDA Version: 11.5

I appreciate your help. Thanks!

regards,
Gero

[Limbicnation]

When I follow these instructions:

conda create -n ASMA python=3.7 conda activate ASMA conda install pytorch==1.8.1 torchvision==0.9.1 torchaudio==0.8.1 cudatoolkit=11.3 -c pytorch -c conda-forge pip install pyyaml paramiko pandas requests tensorboard tensorboardX tqdm

I get the following error:

anaconda3/envs/ASMA/lib/python3.7/site-packages/torchvision/transforms/functional.py", line 45, in _interpolation_modes_from_int return inverse_modes_mapping[i] KeyError: 1920

[birolkuyumcu]

for KeyError @Limbicnation

open "ASMAGAN/data_tools/test_data_loader_resize.py"
at line 36 correct like this
transform.append(T.Resize((1088,1920)))

[Limbicnation]

@birolkuyumcu Thank you very much!
This works perfectly now!
👍