FULL STORY

PyTorch Conference Debuts in China

PyTorch Conference was held in China for the first time, co-located with KubeCon + CloudNativeCon + OpenInfra Summit in Shanghai on September 7-9.

2026-09-01 ~ 2026-09-09 · 2 episodes · 10 posts

Episode 1 · KubeCon and PyTorch Conference Jointly Held in Shanghai This September (2026-09-01, 2 posts)

KubeCon, CloudNativeCon, OpenInfra Summit and PyTorch Conference will jointly take place in Shanghai on September 7–9, with Huawei as a strategic sponsor announced by PyTorch.

Episode 2 · PyTorch Conference China 2026 Debuts in Shanghai Alongside KubeCon (2026-09-07, 8 posts)

PyTorch Conference China 2026 took place in Shanghai on September 7-9, 2026 — the first PyTorch conference ever held in China — co-located with KubeCon + CloudNativeCon and OpenInfra Summit, bringing the PyTorch, Kubernetes and OpenStack communities together. Themed "Open Source in the AI Era," the event was officially expected to draw over 1,000 developers and 400+ participating companies; opening posts confirmed it kicked off as scheduled, and the second day of keynotes concluded successfully.

Confirmed

  • The conference ran September 7-9 in Shanghai, co-located with KubeCon + CloudNativeCon + OpenInfra Summit, marking PyTorch's first conference in China; opening posts (m4/m5) confirmed it opened on schedule.
  • The theme was "Open Source in the AI Era"; officials reported the second day of keynotes concluded, with the team showcasing open-source full-stack AI work spanning frameworks, compilers, inference, hardware and Agents at its booth (m6).
  • Official forecasts projected 1,000+ developers and 400+ participating companies.
  • Ahead of the event, PyTorch Foundation Executive Director sparkycollier appeared at the Shanghai FlagOS-hosted "Open Computing: Building Open Software for Diverse Hardware" event, noting PyTorch's 80 million monthly downloads and discussing open-source software stacks for diverse hardware (m1).
  • Huawei hosted a dedicated session covering cross-accelerator workloads and PyTorch adaptation in multi-chip environments (m2/m8).

Why it matters

  • The first PyTorch conference in China, co-located with flagship cloud-native and OpenInfra events, reflects the rising standing of China's developer community and diverse-hardware ecosystem in the global open-source AI stack.
  • Huawei's session on cross-accelerator and multi-chip adaptation echoes the PyTorch community's ongoing push for heterogeneous hardware support, offering a reference point for domestic accelerators integrating into mainstream deep learning frameworks.
  • The booth's full-stack presentation — from frameworks and compilers to inference, hardware and Agents — shows the community's focus expanding from training frameworks toward end-to-end open-source AI infrastructure.