Mne-python: I've a problem with "plot_sensors_connectivity" visualization

[Sirabhop]

My goal is to compute EEG connectivity, so I run this code below in order to compute connectivity.
con, freqs, times, n_epochs, n_tapers = spectral_connectivity(epoch, method='pli', mode='multitaper', sfreq=2048, fmin=4, fmax=8, faverage=True, mt_adaptive=False, n_jobs=1)

And, I also need to visualize the connectivity, so I use this code
plot_sensors_connectivity(epoch.info, con[:, :, 0])
But, turn out it gave me this error

AttributeError: 'NoneType' object has no attribute '_Renderer'

(I have used EEG biosemi32 measuring brain wave, set montage with 'biosemi32')
epoch.set_montage('biosemi32')

Thank you

[GuillaumeFavelier]

Hello @Sirabhop and thank you for your reporting your issue.

Please could you share the output of:

python -c "import mne; mne.sys_info()"

Plus maybe a minimal example to reproduce? It would be really helpful.

[Sirabhop]

mne.sys_info()

Platform: Linux-4.19.104+-x86_64-with-Ubuntu-18.04-bionic
Python: 3.6.9 (default, Apr 18 2020, 01:56:04) [GCC 8.4.0]
Executable: /usr/bin/python3
CPU: x86_64: 2 cores
Memory: 12.7 GB

mne: 0.20.4
numpy: 1.18.4 {blas=openblas, lapack=openblas}
scipy: 1.4.1
matplotlib: 3.2.1 {backend=module://ipykernel.pylab.backend_inline}

sklearn: 0.22.2.post1
numba: 0.48.0
nibabel: 3.0.2
cupy: Not found
pandas: 1.0.3
dipy: Not found
mayavi: Not found
pyvista: 0.24.2
vtk: 8.1.2

[GuillaumeFavelier]

I see that you have pyvista installed and not mayavi. Could you also try:

python -c "import pyvista; print(pyvista.Report())"

[Sirabhop]

I got session crashed

ERROR:root:bad X server connection. DISPLAY=

Is it about my computer? because I run all my code via Google Colab

[GuillaumeFavelier]

ERROR:root:bad X server connection. DISPLAY=

AFAIK, this means that the environment does not provide a window with an opengl context. I'm afraid this is an implicit requirement for now with the pyvista 3d backend (it's more a vtk issue really).

A way to mitigate could be to use a xvfb server to simulate this but I'm not familiar enough with Google Colab to help you set up that. A config example is available in .circleci/config.

On the bright side, there is active work to add support for headless remote visualization as we speak in https://github.com/mne-tools/mne-python/pull/7758 for example.

Any other ideas to make it work @agramfort , @hoechenberger ?

[GuillaumeFavelier]

Little update on this. With #7758 merged, there is now a "notebook" 3d backend available! This has been tested to work as expected on Jupyter Notebook.

About support for Google Colab, I tried setting it up with @drammock without much success so far. Mainly because we require mesalib from conda and configuring conda is not straightforward in this scenario...

FYI, I plan to move on to Jupyterlab support next.