Upgrading the CUDA version in Kaggle notebook

I’m trying to upgrade the CUDA version in a Kaggle notebook using a custom Docker image. Can anyone help me by providing the steps I need to follow to successfully complete this process?