Building Tensorflow on Jetson Xavier, Jetpack 4.2.1

Robotics & Edge Computing Jetson Systems Jetson AGX Xavier
1

I based on this topic:

My version of Jetpack: Jetpack 4.2.1
Python: Python 3.6.8

My version of Bazel:
INFO: Invocation ID: 689808cc-92f6-41f9-b49b-dc4b74e1c25c
Build label: 0.21.0- (@non-git)
Build target: bazel-out/aarch64-opt/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar
Build time: Tue Jul 30 13:28:46 2019 (1564493326)
Build timestamp: 1564493326
Build timestamp as int: 1564493326

My version of TF: tensorflow-1.13

I did all instruction, but i got next errors:

ERROR: Skipping ‘//tensorflow:libtensorflow_cc.so’: error loading package ‘tensorflow’: in /home/jetpack/Workspace/tensorflow-1.13/tensorflow/tensorflow.bzl: Unable to load package for ‘@local_config_rocm//rocm:build_defs.bzl’: The repository could not be resolved

WARNING: Target pattern parsing failed.

ERROR: error loading package ‘tensorflow’: in /home/jetpack/Workspace/tensorflow-1.13/tensorflow/tensorflow.bzl: Unable to load package for ‘@local_config_rocm//rocm:build_defs.bzl’: The repository could not be resolved

INFO: Elapsed time: 0.338s
INFO: 0 processes.

FAILED: Build did NOT complete successfully (0 packages loaded)
currently loading: tensorflow
Fetching @local_config_cuda; fetching

How can i solve the problem?

2

do you have network connectivity?

3

Yes i have.

4

seems like the repo location may have changed then - i’m not enough of an expert to help beyond that…but maybe see if the locations are still valid

5

I am not sure that the problem related with location.

To make sure this,
I did same things with Bazel 0.18.1 + TF 1.9, and build run without errors.

6

Hi,

It looks like that the rocm url is somehow broken.

ERROR: error loading package 'tensorflow': in /home/jetpack/Workspace/tensorflow-1.13/tensorflow/tensorflow.bzl: Unable to load package for '@local_config_rocm//rocm:build_defs.bzl': The repository could not be resolved

Would you mind to turn it off and try it again?

Do you wish to build TensorFlow with ROCm support? [y/N]:<b> n</b>
No ROCm support will be enabled for TensorFlow.

Here is an experiment sharing for building TensorFlow v1.13 on Xavier for your reference:
https://devtalk.nvidia.com/default/topic/1055131/jetson-agx-xavier/building-tensorflow-1-13-on-jetson-xavier/

Thanks.

7

Hello,

ROCm is turn off.

The line from .tf_configure.bazelrc

build --action_env TF_NEED_ROCM="0"
8

Hi,

But it looks like that bazel still try to access the repository.
Could you try the configure shared in the topic 1055131 and start a clean build?

Thanks.

9

Hello AastaLLL,

I have reinvented all my steps and did clean install everything.
Now all works perfectly.

Thanks.

10

Good to know it works now.
Thanks for updating the status with us : )