Building tensorflow .whl from source for Jetson AGX (Solved)

For my work, I need to build tensorflow with a different version(r1.11), the current .whl files are for 1.10(https://developer.download.nvidia.com/compute/redist/jp40/tensorflow-gpu/). Are these built using the standard tensorflow way(https://www.tensorflow.org/install/source), or should I be aware of some optimizations etc that were used when building them

Hi,

For the official package, it is built from our internal TensorFlow branch and goes through detail testing:
https://devtalk.nvidia.com/default/topic/1042125/jetson-agx-xavier/official-tensorflow-for-jetson-agx-xavier/

For the community package, it is built directly from the public TensorFlow branch and the test is limited.
https://devtalk.nvidia.com/default/topic/1041712/jetson-agx-xavier/tensorflow-wheel-for-jetpack-4-0-/

You can check this GitHub for the patch to build TensorFlow for Jetson.
Suppose the modification on v1.11 should be similar.

Thanks.