You are an amazing machine learning researcher. You invented a new super cool method. You are not sure that it is significantly better than your baseline. You don't have 3000 GPUs to rerun your ...