Building TensorFlow v1.4.1 from source on virtual environment created with python 2.7 on jetpack 4.5.1

Hi, I currently have tensorflow program code in python 2.7 for image detection on our old Jetson Tx2. Now I am working with new Jetson XAVIER NX and want the application to run similarly on it, but i am facing some issues.
First I will explain how i have setup my jetson machine.
1.Jetpack 4.5.1 is installed on jetson.
2.python 3 and tensorflow v1.15 is installed.
3.virtual environment containing python 2.7 is installed.
Following is the problem I am facing while running the code.
1.Since my code is in python 2.7 so python3 and tf v1.15 I have installed is not useful.
2.Now option remaining with me is to code in my virtual environment which is with python 2.7 but tensorflow package for python 2.7 is stopped by nvidia for jetpack 4.5.1.
3.Then it is suggesting to build package from source according to this nvidia forum link-Install tensorflow 1.15 for Python 2.7
4.While building tensorflow v1.4.1 I did following things.
1.created virtual environment with python 2.7 and activate it.
2.install pip, jdk8.
3.Added bazel installation to package source.
4.Installed tf v1.41 compatible bazel v0.9.0.
5.cloned and configured tensorflow successfully.
Now when i try to do build tensorflow using following cvommand i get error and tensorflow is not built.
command :- bazel build --config=opt --config=cuda --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" //tensorflow/tools/pip_package:build_pip_package.
Error message:- exec format error.
so am unable to build tensorflow from source code for python 2.7.

It will be really helpful if provided one of the following things.
1.official version of tensorflow supporting python2 for Jetpack 4.5.1
2.Step wise guidance regarding building tf 1.4.1 from source for python2 on ubuntu 18.04.

Hi,

Unfortunately, we stop to release the TensorFlow package for python2.7 for a while.
Would you mind migrating to python3.6?

Thanks.

1 Like

Hi,

Thanks for the prompt replay.
We are facing issues migrating from 2.7 to python 3.9 I am using future library any help will be appreciated.

Thanks.

Hi,

Maybe you can try to build TensorFlow v1.15.x for python2.
Here is an example for your reference:

Thanks.

1 Like