Flash TX2 stuck at 99.9%

I’ve been flashing quite a few TX2s over the last couple of months, but lately -after upgrading to SDKMANAGER 1.0, I’ve experienced an issue quite a few times.

During the initial flashing of a completely new TX2, I got this error:

TensorFlow : ERROR: tensorboard 1.14.0 has requirement setuptools>=41.0.0, but you'll have setuptools 39.0.1 which is incompatible.

I fixed this by sshing into the TX2, and

apt install python3-testresources
pip install --upgrade setuptools

After I restart the flash process, and only select Jetson SDK components and Additional SDKs, it slows down and ends up stopping completely after a few hours at 99.9% -during the installation of Tensorflow.

The third time I restart the sdkmanager, now only selecting to install Tensorflow, it immediately completes with “successful installation” with the status of Tensorflow says “download pending”.

Both the host computer and the TX2 are online, and everything is connected. All other steps complete successfully.

The computer is native Ubuntu Linux 18.04.

Is there a way I can install tensonflow from the TX2 with apt or pip?


Which TensorFlow package do you try to install?

There are three different packages as version v1.14.0.
You will need to install the one built with the same JetPack environment as your system.

So it’s recommended to flash/install the package at the same time.
As there are some newer release, would you mind to give JetPack4.3+TensorFlow v1.15.0 a try?


Hi there!

I am trying to install the Tensorflow that is part of the sdkmanager for Jetson 4.2.2. I didn’t pay attention to which version that is. But you would know that, right?

Unfortunately, I can’t upgrade to 4.3 now, as we have some unresolved dependendencies to some Leopard Imaging drivers that are compiled for 4.2.2 that have to run on that TX2.

I do flash and install packages at the same time, but as the sdkmanager ends up hanging, and there is no need to repeat the whole 3hours (!) process again, I restart the sdkmanager only selecting the additional SDKs to install.

Can I install these files? https://developer.download.nvidia.com/compute/redist/jp/v42/tensorflow-gpu/ I’m guessing that jp/42 means Jetson Pack 4.2?


Sorry for the delay.

Yes. The package listed int eh link is for JetPack4.2.
It is compatible for the JetPack4.2.