CANLI
xAI, Imagine API’yi 2.0’a Yükseltmeye Hazırlanıyor: Görüntü ve Video Tek…·Microsoft MAI-Cyber-1-Flash’ı Duyurdu·Moonshot AI, Kimi K3 Model Ağırlıklarını ve Teknik Raporunu Açık…
12 Nov 2026 · 19:22 GMT+3
Ai Haber – Türkiyenin Yapay Zeka Haber Portalı
ARAşTıRMA · YAPAY ZEKA arXiv:2610.24161 16 Eki 2026 · v3

TorchTPU – A Hybrid Eager/Compiled Architecture for Native PyTorch Execution on TPU

Claudio Basile, Michael Voznesensky, Gunhyun Park, Ben Wilson, Parker Holloway, +48 yazar

YAYIN:16 Eki 2026 ALAN:cs.AI OKUMA:10

Özet

TorchTPU is an open-source collaboration between Google and Meta that introduces a native and principled PyTorch backend for Google TPUs. It enables seamless out-of-the-box model execution and TPU migration while delivering industry-competitive inference and training performance through eager execution optimizations, torch.compile integration, and native hardware acceleration.

Özetle: TorchTPU is an open-source collaboration between Google and Meta that introduces a native and principled PyTorch backend for Google TPUs.

Özet

TorchTPU is an open-source collaboration between Google and Meta that introduces a native and principled PyTorch backend for Google TPUs. It enables seamless out-of-the-box model execution and TPU migration while delivering industry-competitive inference and training performance through eager execution optimizations, torch.compile integration, and native hardware acceleration.

Orijinal Özet (İngilizce)

TorchTPU is an open-source collaboration between Google and Meta that introduces a native and principled PyTorch backend for Google TPUs. It enables seamless out-of-the-box model execution and TPU migration while delivering industry-competitive inference and training performance through eager execution optimizations, torch.compile integration, and native hardware acceleration.

Kaynak: arXiv:2610.24161 · PDF

BibTeX

@article{basile2026torchtpu,
  title   = {TorchTPU - A Hybrid Eager/Compiled Architecture for Native PyTorch Execution on TPU},
  author  = {Claudio Basile and Michael Voznesensky and Gunhyun Park and Ben Wilson and Parker Holloway and Zhanyong Wan and Vlad Belous and Manish Khanna and Paul Ganssle and Kevin Gleason and Andrew Dame and Brendan Folie and Bhavya Bahl and Haifeng Jin and Pedro Goncalves Mokarzel and Arian Arfaian and Luke Boyer and Chi Zhang and Yiling Yuan and Adam Mainz and Aditya Patel and Huangze Tang and Jay Gan and Lizzie Liang and Will Froom and Tushar Agarwal and Karlo Basioli and Theotime Combes and Emilio Cota and Minos Papadopoulos and Tudor Pipernea and Yarong Mu and Qi Zhou and Yaoshiang Ho and Aleksey Vlasenko and Alexander Brauckmann and Ziyin Huang and Jack Oh and Jerry Hsu and Tanvi Bajpai and Jialei Chen and Wenjun Gu and Chintan Shah and Chris Chan and Sayce Falk and Chris Jones and Rob Torres and Kat Ko and Lee Howes and Freya Azad and Andi Gavrilescu and Joe Pamer and Bill Jia},
  journal = {arXiv preprint arXiv:2610.24161},
  year    = {2026},
  url     = {https://arxiv.org/abs/2610.24161}
}

Tartışma

Bu habere emoji ile tepki ver

Hizli:

Henüz yorum yok. İlk yorumu siz yapın!

Yapıcı ve saygılı yorumlar bekliyoruz. Topluluk kuralları