CUDA RuntimeError: Cudnn_Status_Not_Initialized

When your Python application encounters the ‘cudnn_status_not_Initialized’ error, you should first check your Python application’s CUDA version. If this is not the case, you can manually force cuDNN to initialise by using a mock convolution, which ignores the GPU’s limit.

If the error persists after this, your problem is most likely due to an outdated version of CuDNN. It is important that you check the latest version of your CuDNN image and its Docker image. Also, be sure to include the –nv flag, which lets the container use the NVIDIA driver installed on your system.