The torch.profiler tool enables developers to track CPU and GPU execution times across model iterations. This guide explains how to identify bottlenecks and optimize memory usage during training. It provides a practical starting point for developers. Proper profiling reduces compute waste and speeds up inference for PyTorch users.