Profiling in PyTorch (Part 1): A Beginner's Guide to torch.profiler

Profiling is essential for optimizing PyTorch models, but its steep learning curve can be a significant barrier for developers. The `torch.profiler` tool helps demystify performance by tracing the entire chain of events, from a Python function call down to the individual CUDA kernels running on the GPU. This guide makes profiling approachable by starting with a simple matrix multiplication and addition, teaching you how to read the profiler's output and identify performance characteristics. Ultimately, you will learn how to interpret CPU and GPU activity and see what changes when optimizations like `torch.compile` are introduced.