Deeplabcut: Error when importing DeepLabCut on AWS

[N-Sensho]

Hi,
I'm trying to use DLC on AWS, but having this error:
ImportError: Cannot load backend 'TkAgg' which requires the 'tk' interactive framework, as 'headless' is currently running

I'm using Deep Learning AMI (Ubuntu) Version 24.0, and GPU instance type p2.xlarge. Ubuntu version is 16.04.6 LTS.

What I did after connecting to the instance is:

conda create -n behavior python=3.6
source activate behavior
pip install deeplabcut
pip install https://extras.wxpython.org/wxPython4/extras/linux/gtk3/ubuntu-16.04/wxPython-4.0.3-cp36-cp36m-linux_x86_64.whl
pip install tensorflow==1.10

This is the screenshot.

I checked that the version of matplotlib was 3.0.3.

Also, in a new environment, I installed tensorflow-gpu instead of tensorflow1.10 and then followed the instruction of this URL( https://www.tensorflow.org/install/gpu ) for Ubuntu 16.04 (CUDA9.0), but got the same error.

I did the same thing in Deep Learning AMI (Ubuntu) Versions (if I remember correctly) 8.0, 12.0, 16.0, 21.0, but got the same error every time.

Does anybody have a solution?

Thank you.

[MMathisLab]

On AWS you will need to run DLC in "light" mode, i.e. without GUIs. Here is how to set that before you import:

https://github.com/AlexEMG/DeepLabCut/blob/master/docs/UseOverviewGuide.md#option-2-using-the-program-terminal-start-ipython

in short, check the the Docker documentation, but export DLClight=True

[AlexEMG]

@N-Sensho does that fix the issue on AWS? Otherwise check out variants for changing the backend, e.g. https://www.google.com/search?q=cannot+load+backend+%27tkagg%27+which+requires+the+%27tk%27+interactive+framework

[N-Sensho]

I'm sorry for the late reply.
It worked! Thank you!