Hi,
I'm trying to cross-compile TensorFlow Serving on an x86_64 host for Jetson Nano (aarch64 target) without using Docker or any container. However, the process with Bazel (hermetic 💩) has been extremely complex and fragile, with many issues related to toolchain setup and platform configuration.

Is there any official or community-supported guidance for this kind of setup? Or a minimal working example that uses a custom toolchain to cross-compile directly?

Thanks!