PyTorch Day Japan 2026 Will Convene in Tokyo on Dec 10

The PyTorch Foundation, in partnership with Hugging Face, IBM, and Mitsubishi Electric, will host PyTorch Day Japan 2026 in Tokyo on Thursday, December 10. This single-day technical gathering aims to unite machine learning engineers, researchers, and open-source enthusiasts. The event will focus on knowledge sharing and collaborative development across the broader AI ecosystem, highlighting both core library updates and foundation-hosted projects.

Attendees will explore a variety of critical open-source technologies, including vLLM, DeepSpeed, Ray, Helion, and Safetensors. The conference program is structured around three primary tracks. The first, Sovereign AI and local open models, addresses strategies for running open-weights systems on local infrastructure, covering post-training, adaptation, and privacy-first deployments. The second track focuses on physical and edge AI, showcasing how to deploy models on robotics and hardware using vision-language-action models and real-time optimization. The final track covers the PyTorch ecosystem itself, highlighting TorchVision, TorchAudio, TorchRL, PyTorch Distributed, and compiler technologies like torch.compile in PyTorch 2.x.

The call for proposals is currently active, with the submission deadline set for Sunday, September 27 at 11:59 PM JST. Organizers will announce the final schedule on Wednesday, October 14, and selected speakers must submit their presentation slides by Wednesday, December 9. Registration is also open, featuring an early-bird rate of ¥5,000 until November 11 at 11:59 PM JST, which offers a savings of ¥3,000 off the standard price. Discounted academic registration is also available for students and faculty members.

For machine learning practitioners, this event provides a vital opportunity to align local development strategies with the latest upstream optimizations in PyTorch 2.x. By learning directly from the creators of tools like DeepSpeed and vLLM, developers can better navigate the complexities of deploying resource-constrained edge models and securing private, sovereign AI infrastructure.