Build environment for NVIDIA Jetson series for python/cython apps
83
When deploying non-containerized apps (eg. debian packages) which require Python, its best to use python's virtual environment. However, for off-device compilation, this causes a few issues:
venv does not contain actual Python, so for it to work, host must have exactly the same version of Python, with which venv was created. Easiest way to insure that, is to build package on the same base OS, in case of Jetpack 5.X that is ubuntu:focal.numpy, pysoem) must be architecture-specific. In case of Jetson, that is aarch64/arm64v8. However, running it on amd64 host requires qemu, which makes things considerably slow.For these reasons, its better to have a ready docker image with pre-installed dependencies required for creating Python's virtual environment with required version of Python, arm64v8, and gcc for Cython compilation.
Content type
Image
Digest
sha256:5c75ccbfd…
Size
152.4 MB
Last updated
about 3 years ago
docker pull anynameaaa/arm64v8_buildenv:focal