Jupyter kernel dying when trying to run conv neural network code in wsl with h/w acceleration

Question

I am doing basic code of classifying cats and dogs using convolutional neural network. I have 4000 images of dogs and 4000 images of cats. When the code goes to the part where we train the model the kernel dies. I have wsl on windows and I am doing operations after I have run the command 'conda create --name tf python=3.9' and 'conda activate tf' Versions: tensorflow==2.13, cuda==11.8.0, cudnn==8.6.0.163 (followed steps from official tensorflow page and confirmed working) My GPU is Nvidia RTX 3060 (notebook/laptop GPU)

I tried to uninstall/install Jupyter notebook.

I tried limiting the memory with " gpus = tf.config.experimental.list_physical_devices('GPU') for gpu in gpus: tf.config.experimental.set_memory_growth(gpu, True)"

I am hoping that epochs will run and model will get trained with help of GPU acceleration.

Answer 1

Well the problem got solved when I downgraded my cuda and cudnn versions. So in my virtual environment I have python=3.9, tensorflow=2.10, cuda=11.2 and cudnn=8.1.0.