You are training a computer vision model to identify the type of government ID in images, using a GPU-powered virtual machine on Google Compute Engine. The training parameters include: Optimizer : SGD , Image shape : 224x224 , Batch size : 64 , Epochs : 10 , Verbose : 2. However, you encounter a 'ResourceExhaustedError: Out Of Memory (OOM) when allocating tensor' during training. What steps should you take to resolve this issue?