Sponsored Session: Empowering AI Everywhere: Democratizing PyTorch with Intel...
summary
PyTorch runs natively on Intel AIPCs using an XPU device with >90% ATen ops implemented (via SYCL-like kernels and oneDNN/oneCCL backends), a Triton-enabled inductor path and a template-kernel library in progress. Torch.compile is supported on Windows — torch.compile on Windows — and Executor 1.0 flows to OpenVINO for heterogeneous CPU/GPU/NPU execution. LM inference uses low-precision formats and fused attention kernels to cut memory needs and boost throughput; demos and benchmarks show significant speedups using the torch_xpu package.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 3:41PyTorch on Intel GPUs
- 5:53Intel GPU support pillars
- 16:50LLM on Intel iGPUs
- 19:59Library and ecosystem
- 22:30Roadmap and call
- 24:14Q&A
related talks
24mIntel and Open Source: Bringing AI Everywhere
39mRunning On-Device AI With Qualcomm AI Engine Using LiteRT & ExecuTorch
24mIndustry Panel
01h 19mWORKSHOP 2: AI FOR EVERYONE
22mLeverage AI Kits for Faster AI Solutions Tailored to Your Needs
06mKeynote: PyTorch Lightning: Powering the GenAI Revolution from Research to the...
11mWhy Windows is the best place for AI development
37mEnabling AI For Everyone - Hiai Open Platform Introduction