Request for the new updates on the tutorials on the google colab

[linlupin]

As mentioned by Tyler, google colab had some updates on Jan. 13th while Python reminded the same version.
It causes the conflict using some old packages or bugs to install the old packages.
After checking the contents of the old tutorials, I would like to request the following updates to run the tutorials in the new environment of Google Colab:

• Day_1/Tuto_1.2

pip install -q 'gwpy==3.0.8' 
>>> pip install -q 'gwpy==3.0.11’

• Day_1/Tuto_1.3

pip install -q 'gwpy==3.0.8' 
>>> pip install -q 'gwpy==3.0.11’

plot.refresh() has the bug, so I suggest to mark it.

plot.refresh()  >>> # plot.refresh()

• Day_1/Tuto_1.4

! pip install -q PyCBC==2.4.0 lalsuite==7.21 
>>>  ! pip install -q PyCBC==2.4.1 lalsuite==7.21

• Day2/Turo_2.1-2.3
  To avoid the warning messages, the following code is required to add:

import warnings
warnings.filterwarnings("ignore", "Wswiglal-redir-stdio")
import lal

• Day3/Tuto_3.1
  Installation of the old packages will show the message "plotnine 0.14.5 requires matplotlib>=3.8.0"

! pip install -q bilby==2.2.2 matplotlib==3.7.3 dynesty==2.1.2 corner==2.2.2
>>> ! pip install -q bilby==2.4.0 matplotlib==3.8.0 dynesty==2.1.2 corner==2.2.2

• Day3/Tuto_3.2

! pip install -q bilby==2.2.2 matplotlib==3.7.3 dynesty==2.1.2 corner==2.2.2 gwpy==3.0.8 lalsuite==7.21
>>> ! pip install -q bilby==2.4.0 matplotlib==3.8.0 dynesty==2.1.2 corner==2.2.2 gwpy==3.0.11 lalsuite==7.22

import warnings
warnings.filterwarnings("ignore", "Wswiglal-redir-stdio")
import lal

• Day3/Tuto_3.3

! pip install -q 'corner==2.2.2' 'bilby==2.2.2' 'astropy==6.0.1'
>>> ! pip install -q 'corner==2.2.2' 'bilby==2.2.2' 'astropy==6.0.1' 'matplotlib==3.8.0'

• Tuto_A.1

! pip install -q lalsuite==7.21 pesummary==1.2.0
>>> ! pip install -q lalsuite==7.22 matplotlib==3.8.0 pesummary==1.3.4 

In the Initialization, the following code is required to avoid the warning message:

import warnings
warnings.filterwarnings("ignore", "Wswiglal-redir-stdio")
import lal

• Tuto_A.3
  The following code needs to be claimed again:

import warnings
warnings.filterwarnings("ignore", "Wswiglal-redir-stdio")
import lal

There is an error to plot the color bar though the contour map can still be presented.
The error message is like the following:

Error to plot the color bar
pyhough.pm.python_plot_triplets((flat_times-flat_times[0])/86400,flat_freqs,flat_pows,'.',label='equalized power')
++++++++++++++++
ValueError: Unable to determine Axes to steal space for Colorbar. Either provide the *cax* argument to use as the Axes for the Colorbar,
provide the *ax* argument to steal space from it, or add *mappable* to an Axes.

I have tried installing the pyhough and matplotlib of the latest version, but this error still remains.
It seems the error is caused by the pyhough package, and I have no idea to correct it since I am not familiar with it.

Since Google Colab may update its environment again, I suggest checking all the tutorials one week before the workshop again.

[duboism]

Please see #10

[duboism]

Hello,

I opened #22 to solve this.

@linlupin can you try to download the modified notebooks (https://github.com/duboism/odw/archive/refs/heads/reupdate_environment.zip) and run them on Colab ? According to my tests, they run (except Extention Topic A.2 but I think it's a CUDA error on Colab).

[linlupin]

Hello,

I opened #22 to solve this.

@linlupin can you try to download the modified notebooks (https://github.com/duboism/odw/archive/refs/heads/reupdate_environment.zip) and run them on Colab ? According to my tests, they run (except Extention Topic A.2 but I think it's a CUDA error on Colab).

Dear Mathieu,

I have checked all the tutorials on the Google Colab including the extension topic.
It is a little weird that I cannot see the LVK logo.
It might be because I downloaded the files on my Google Drive and then checked it while the path of the LVK ligo should be changed.

As you mentioned, the Tuto_A.2 has the problem on this section:
likelihood = gwpopulation.hyperpe.HyperparameterLikelihood(posteriors = posteriors, hyper_prior = population, selection_function = p_det)
The error message is the CUDARuntimeError: cudaErrorInsufficientDriver: CUDA driver version is insufficient for CUDA runtime version.
It seems to be caused by cupy package; however, I run the code with CPU only and did not use GPU.
I did not meet this bug a month ago, and this error is strange for me.

For the Tuto_A.2, I noticed that the 1st and 3rd figures have demonstrated weaker power than those presented in the original/old temporary plots.
Though the curve in the figure passed the same trajectory, I cannot ensure whether it is indeed correct.

Except for what I mentioned, I suggest all other scripts are fine.
Nevertheless, I still suggest re-checking them one more time with 1 or 2 weeks just before the workshop to confirm that all the participants can run tutorials without bugs.

[duboism]

Thanks @linlupin !

For the logo, I think I accidentally removed it. I will have a look at that ASAP.

For the CUDA/cupy problem, I don't know what to do. I asked a question on Stack Overflow: https://stackoverflow.com/questions/79493715/using-cupy-on-google-colab. Let's see if we make progress.

[duboism]

@linlupin, for the logo I think you have the correct explanation: they are not displayed on Colab because it doesn't access the logo (I was confused because I had a local issue but it doesn't impact Colab or anything else). There is nothing we can do about that.

[duboism]

I have no response on stack overflow so I opened a bug report on Google Colab's GitHub: googlecolab/colabtools#5166.

[duboism]

After discussion with the Colab team, we think the problem comes from cupy. They offered a workaround: switch to a GPU runtime. I will add this on the notebook A.2.