events
GPU Detective Stories: PyTorch Recompiles & Hidden Bottlenecks
An online session from the AI Performance Engineering Meetup in Boston, hosted by Chris Fregly and Antje Barth, with two technical talks on hidden costs in AI workloads.

En inglés
- Starts
- 2026-11-16 17:00 UTC
- Organizer
- AI Performance Engineering Meetup (Boston)
- Location
- Online event
An online session from the AI Performance Engineering Meetup in Boston, hosted by Chris Fregly and Antje Barth, with two technical talks on hidden costs in AI workloads. Chaim Rand compares six ways to handle PyTorch recompilation during inference, from cache preloading and bounded recompilation with eager fallback to ahead-of-time compilation with torch.export and AOTInductor, discussing how model, input distribution, runtime and latency shape the choice. Israel Ogbole of Zymtrace walks through a case study on training and serving the model Jev, using continuous CPU/GPU profiling to find compute, memory, communication and synchronization bottlenecks, and the measured gains after each change.
Event website