While reading the "Profiling in PyTorch" blog series on @huggingface, I found an efficient setup: running the @PyTorch profiler on @modal without a local GPU. The traces save locally, so you can drop them directly into Perfetto to analyze. Link to the GitHub repo in the comments!
0:16