Jetson nano 2GB and tensorflow not really usable

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

I finally managed to get tensorflow working thanks to a great helpful article
The problem is 2GB of memory is just not enough and its way slower than a laptop with a regular cpu 6 way core

Simple mnist tests run out of memory almost immediately. This was bought just for CUDA learning and AI learning.

I am about to spend more money on a Jetson nano orin 8GB. I hope I am not going to have similar problems.

if so please let me know which Jetson has enough memory to be useful

Regards
Laurence

2

I meant to add. I am in runlevel 3 no GUI and using ssh.

The helpful article :

4

Hi,

Orin Nano has much more capacity and below is the model benchmark for your reference:

https://developer.nvidia.com/embedded/jetson-benchmarks

Jetson Pretrained Model Benchmarks
> Jetson Orin Results
> Jetson Orin Nano 8GB

TensorFlow requires much more memory to execute.
To save memory, you can also try to convert the model into ONNX format and deploy it with TensorRT.

Thanks.

5

Thank you
I ordered my Orin yesterday.
Regards
Laurence

6

Hello

I have my Orin Nano, the Jetpack it installed is jetpack-runtime_6.0-b52_arm64
I cannot seem to find a way to get tensorflow installed

I did this:

sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev libblas-dev gfortran

sudo apt-get install python3-pip
sudo python3 -m pip install --upgrade pip
sudo pip3 install -U numpy==1.22 future==0.18.2 mock==3.0.5 keras_preprocessing==1.1.2 keras_applications==1.0.8 gast==0.4.0 protobuf pybind11 cython pkgconfig packaging h5py==3.7.0

sudo pip3 install --extra-index-url https://developer.download.nvidia.com/compute/redist/jp/v$JP_VERSION tensorflow==$TF_VERSION+nv$NV_VERSION

But the last part failed with

Looking in indexes: Simple index, https://developer.download.nvidia.com/compute/redist/jp/v
ERROR: Could not find a version that satisfies the requirement tensorflow==+nv (from versions: 2.10.0rc0, 2.10.0rc1, 2.10.0rc2, 2.10.0rc3, 2.10.0, 2.10.1, 2.11.0rc0, 2.11.0rc1, 2.11.0rc2, 2.11.0, 2.11.1, 2.12.0rc0, 2.12.0rc1, 2.12.0, 2.12.1, 2.13.0rc0, 2.13.0rc1, 2.13.0rc2, 2.13.0, 2.13.1, 2.14.0rc0, 2.14.0rc1, 2.14.0, 2.14.1, 2.15.0rc0, 2.15.0rc1, 2.15.0, 2.15.1, 2.16.0rc0, 2.16.1)
ERROR: No matching distribution found for tensorflow==+nv

7

I think I found the correct tensorflow
tensorflow-2.15.0+nv24.02-cp310-cp310-linux_aarch64.whl

8

In case this is useful for others

So complete steps to make both torch and tensorflow work on the 8GB Orin nano

I flashed the Orin Nano using sdkmanager
I then imaged a 512G microSD and also installed and NVME 250G device
Completed the install and updated Ubuntu
Created an EXT4 on the nvme for user files for AI testing

Note the image I used installed jetpack-runtime_6.0-b52_arm64 and is
jp60dp-orin-nano-sd-card-image.zip

sudo apt-get install libhdf5-serial-dev hdf5-tools libhdf5-dev zlib1g-dev zip libjpeg8-dev libblas-dev gfortran

sudo apt-get install python3-pip
sudo python3 -m pip install --upgrade pip
sudo pip3 install -U numpy==1.22 future==0.18.2 mock==3.0.5 keras_preprocessing==1.1.2 keras_applications==1.0.8 gast==0.4.0 protobuf pybind11 cython pkgconfig packaging h5py==3.7.0

Then copied the torch and tensorflow versions on to the nano
tensorflow-2.15.0+nv24.02-cp310-cp310-linux_aarch64.whl
torch-2.3.0a0+ebedce2.nv24.02-cp310-cp310-linux_aarch64.whl

Ran
sudo pip3 install /new_home/torch-2.3.0a0+ebedce2.nv24.02-cp310-cp310-linux_aarch64.whl
sudo pip3 install /new_home/tensorflow-2.15.0+nv24.02-cp310-cp310-linux_aarch64.whl

Now it is all working for testing

9

Hi,

Thanks for sharing.
Below is the document for TensorFlow and PyTorch for your reference:

Thanks.

12

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.