How to install specific tensorflow version in Jetson Tx2?

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2
, ,
1

Hi I am trying to install Tensorflow version 1.15.3 from Nvidia TensorFlow Container Version 20.09 in Jetson Tx2 for Jetpack 4.5.1 TensorRT 7.1.3. Here are some references which says that above configuration is Tensorflow Container

Both the commands throws error in my terminal

  1. I tried both of these commands
    sudo pip3 install --extra-index-url Index of /compute/redist/jp/v45 tensorflow-gpu==1.15.3+nv20.9

    Looking in indexes: Simple index, Index of /compute/redist/jp/v45
    ERROR: Could not find a version that satisfies the requirement tensorflow-gpu==1.15.3+nv20.9 (from versions: none)
    ERROR: No matching distribution found for tensorflow-gpu==1.15.3+nv20.9

  2. sudo pip3 install --extra-index-url Index of /compute/redist/jp/v45 tensorflow-gpu==1.15.3+nv20.09

    Looking in indexes: Simple index, Index of /compute/redist/jp/v45
    ERROR: Could not find a version that satisfies the requirement tensorflow-gpu==1.15.3+nv20.9 (from versions: none)
    ERROR: No matching distribution found for tensorflow-gpu==1.15.3+nv20.09

Commands are based on this command reference

Support Matrix
TensorRT 7.1.3 Guide
Tensorflow Jetson Release Notes

Some of the Documents states that it supports Tensorflow 1.15.2 while some matrix references states that Tensorflow 1.15.3 can be installed from container 20.09. It would be great and helpful if someone can guide me through this so that I can continue my learning.

2

Hi Virat_Baahubali,

Sorry for the late response, is this still an issue to support?

Thanks

3

I too cant install it.
This is what I ran.

pip3 install --pre --extra-index-url Index of /compute/redist/jp/v45 tensorflow

This is the output

Looking in indexes: Simple index, Index of /compute/redist/jp/v45
ERROR: Could not find a version that satisfies the requirement tensorflow (from versions: none)
ERROR: No matching distribution found for tensorflow

4

Hi Kayccc sorry for the late response basically I just downloaded tensorflow 1.15.5 with Jetpack 4.5.1 TensorRT 7.1.3 from Container 21.05. It is working well so far though I don’t have Ubuntu 20.04 as per the requirement but I took a risk knowing that it would definitely work.

5

Glad to know you fixed it, thanks for the update.

6

Hi,
It doesn’t seem to a “fix” technically. OP used an image that has TensorFlow inbuilt. So, he avoided installing TF himself. I think it’s worth noting that the following tutorial is not working for installing TF in docker containers on Jetson-nano.

Can someone try to reproduce this?

8

Hi,

We try the installation command:

$ pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 tensorflow

And it can run correctly without issue on Nano+JetPack4.5.1.

Please noted that you may meet some issue with the different pip3 installer.
So it’s recommended to use exactly the same package version listed in the below document:

Thanks.

9

Hi AastaLLL,

the command that you mentioned requires Ubuntu 20.04. On my Jetson Tx2 Ubuntu 18.04 is booted though I installed Tensorflow 1.15.5 from container 21.05 for Jetpack 4.5.1 with TensorRT 7.1.3 and its working well so far. Will the OS version make any kind of difference?
Thanks

10

Hi,
Yes, that’s the exact command I executed in my container and it still doesn’t seem to work. I have pasted the contents of my docker file.

FROM  ubuntu

ENV DEBIAN_FRONTEND=noninteractive
RUN apt-get -y update && \
    apt-get -y install iperf3 net-tools iputils-ping gunicorn python3-pip && \
    apt-get -y install tzdata ffmpeg && \
    apt-get -y clean

RUN pip3 install pandas==1.1.5 numpy
RUN pip3 install Flask opencv-python requests pillow

# installing tensorflow
RUN apt-get -y install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev 	\
					zip libjpeg8-dev liblapack-dev libblas-dev gfortran		&& \
	pip3 install -U pip testresources setuptools==49.6.0 					&& \
	apt-get -y install pkg-config libhdf5-dev
RUN	pip3 install -U numpy==1.19.4 future==0.18.2 mock==3.0.5 h5py==2.10.0 	\
					keras_preprocessing==1.1.1 keras_applications==1.0.8 	\
					gast==0.2.2 futures protobuf pybind11
RUN	pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 tensorflow

It is the last step where the docker build fails. It returns the following error.

Step 8/11 : RUN	pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 tensorflow
 ---> Running in eeaf7f0546a1
Looking in indexes: https://pypi.org/simple, https://developer.download.nvidia.com/compute/redist/jp/v45
ERROR: Could not find a version that satisfies the requirement tensorflow (from versions: none)
ERROR: No matching distribution found for tensorflow
The command '/bin/sh -c pip3 install --pre --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v45 tensorflow' returned a non-zero code: 1

Thanks in advance for the help

11

Hi @Virat_Baahubali
How did you use the container image? isn’t it for amd64 and not arm?

12

Installing TensorFlow For Jetson Platform :: NVIDIA Deep Learning Frameworks Documentation Hi @pramodhrachuri you can refer to this link about the installation steps there they have mentioned how to use the container image. Also make sure your jetson is flashed properly with the right version of Jetpack and tensorRT you are trying to install tensorflow for. Most of the Jetson systems have ARM architecture and not amd64.

1 Like