## Silent Broadcasting Can Ruin Your Model PyTorch and TensorFlow tensor broadcasting: how silent shape errors cause difficult-to-debug machine learning bugs Full disclosure: I just wasted ~$4,000 in compute costs last month because of this very silent, very real bug that I've likely been victim to many times over my career and never even knew it. If you are an ML practitioner, or work in deep learning, I can guarantee this has already happened to you, and you most likely never even realized it. It might even be derailing your work right now. In this article I highlight how a single…