events
GPU Detective Stories: PyTorch Recompiles & Hidden Bottlenecks
Due interventi online sulle prestazioni dell’IA: contenere le ricompilazioni di PyTorch durante l’inferenza e analizzare un caso di studio sull’addestramento e l’erogazione di Jev mediante…

In inglese
- Starts
- 2026-11-16 17:00 UTC
- Organizer
- AI Performance Engineering Meetup (Amsterdam)
- Location
- Online event
The AI Performance Engineering Meetup in Amsterdam holds an online session with two talks on hidden costs in AI workloads. Chaim Rand compares six approaches to limiting `torch.compile` recompilation during inference, from cache preloading and warmup seeding to ahead-of-time compilation with `torch.export` and AOTInductor. Israel Ogbole of Zymtrace walks through a case study on training and serving Jev, describing how continuous CPU/GPU profiling located compute, memory, communication and synchronisation bottlenecks. Organisers Chris Fregly and Antje Barth open with introductions and meetup updates.
Event website