PyTorch Conference lands in San Jose on 20-21 October, putting the vLLM, DeepSpeed and Ray teams in one hall
The PyTorch Foundation's North America conference spans the full stack, from inference engines to hardware architectures. Early registration savings end on 4 September.
PyTorch Conference North America 2026, run by the PyTorch Foundation, takes place on 20-21 October at the San Jose McEnery Convention Center.
The programme brings the foundation's projects together: alongside PyTorch, the people working on vLLM, DeepSpeed, Ray, Helion and Safetensors will speak. Added to them are specialists in GPUs and NPUs, model architectures, inference engines, and data centre and edge architectures. Foundation executive director Mark Collier puts the case for gathering the communities this way: new model architectures, hardware architectures and inference engines require coordination between software and hardware people.
Registration is open; tickets start at $249 and the early registration saving ends on 4 September.
The programme and registration are on the event's page.
