Helpful PyTorch Tutorials
September 9, 2026 · IntelliSensei Team
A practical walkthrough of ahead-of-time compiling a PyTorch model into a self-contained shared library with torch.export and AOTInductor: dynamic shapes, the C++ and Python runtimes, packaging, versioning, and when it beats a plain torch.compile server.
Read moreAugust 27, 2026 · IntelliSensei Team
A hands-on walkthrough of taking a PyTorch model on-device: torch.export as the contract, quantization for mobile, lowering to XNNPACK, Core ML and Vulkan delegates, packaging a .pte file, and the failure modes that bite in production.
Read moreAugust 25, 2026 · IntelliSensei Team
The seven failure modes we see most in PyTorch consulting, from eager-mode performance habits and unversioned training data to serving as an afterthought, each with the fix.
Read moreAugust 25, 2026 · IntelliSensei Team
A reproducible benchmark of torch.compile on a ResNet and a small transformer, how to find and fix graph breaks, and when compilation does not pay.
Read moreAugust 25, 2026 · IntelliSensei Team
A recap of the 2.x era (torch.compile, torch.export, ExecuTorch, FSDP2), what died (TorchServe, TorchText), what won (Hugging Face, vLLM, torchao), and what it means for teams planning builds.
Read more