You've set up a Compute Engine instance with an attached GPU, yet the GPU isn't utilized during Tensorflow model training. Which two actions could you take to enable GPU usage for your model training?