Keynote: Fireside Chat with Soumith Chintala, Co-Creator of PyTorch & Merve Noyan

on YouTube ↗32m

" summary

History of PyTorch is condensed into why torch.compile works most of the time and what engineers actually change to make it work — rewrite ops or make graphs CUDA-only — and that PyTorch 2.0 ships a compiler that materially accelerates research workloads. Low‑bit models and compilers reduce cost per inference but overall compute demand keeps rising. TensorFlow + TFX remains easier for production, ONNX is painful, and Hugging Face usage metrics show clear growth in PyTorch adoption.

AI-written summary · about

" chapters (5)

1. 0:00Introduction and background
2. 4:52Origins of PyTorch
3. 8:58Design principles
4. 12:43Scaling and development
5. 19:40Audience Q&A

" related talks

• 48mFireside ChatSoumith Chintala, Ibraham Haddad · Emerging OS Forum @ Open Source Summit Europe 2022
• 54mKeynote: PyTorch 2.1 Technical Deep DiveJoe Spisak, Peng Wu, Mergen Nachin, Mark Saroufim, Yanan Cao, Will Constable, Mario Lezcano · PyTorch Conference 2023
• 24mAdvancements in High-Performance AI/ML through PyTorch's Python CompilerSam Gross, Suraj Subramanian, Justin Jeffress · PyCon US 2023
• 29mKeynoteSoumith Chintala · PyTorch Conference 2022
• 36mDL Compiler Panel DiscussionP. Tillet, J. Ansel, J. Pienaar, T. Chen, M. Zolotukhin, P. Wu · PyTorch Conference 2024
• 12mLightning Talk: State of PyTorchAlban Desmaison · PyTorch Conference 2023
• 12mKeynote: How PyTorch Became the Foundation of the AI RevolutionJoe Spisak · PyTorch Conference 2023
• 15mPyTorch 2.0: Unlocking the Power of Deep Learning with the Torch Compile APIChristian Keller · Open AI + Data Forum @ Open Source Summit North America 2023