After spending 6 hours, building @PyTorch Executorch on Mac to run @Alibaba_Qwen 3 on Android, I documented the full path so others can skip the painful parts: Docker/Linux builds, @Qualcomm_Dev QAIRT/AAR packaging, FastRPC issues using a sample app:
github.com/shivaylamba/ex…
Shivay Lamba@HowDevelop
Working on low-level AI stacks like @PyTorch ExecuTorch or LiteRT-LM for @googlegemma is a good reminder that, even in the age of AI agents, some parts of the developer experience are still surprisingly tricky. The tooling still heavily favors Linux. On Windows, you're often relying on WSL; on macOS, Linux containers become part of the workflow. It all works, but native parity across Linux, Windows, and macOS would make life a lot easier. @carrycooldude would agree 🙃
English
