bump devops tests to include checks for notebook badges

[slayoo]

No description provided.

[AgnieszkaZaba]

I need someone else to rerun demos with numba.

[AgnieszkaZaba]

@slayoo only one notebook should be incorrect: it's performance comparison
Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.

Don't know why the others are wrong, I change them many times. Will try once again.

[slayoo]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.

Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

[AgnieszkaZaba]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.
Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

Ok, you're right. I'm still waiting for Jupyter to finish running :)

[slayoo]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.
Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

Ok, you're right. I'm still waiting for Jupyter to finish running :)

Ah, I didn't realise you are referring to running the GPU tests locally on a machine without GPU to generate notebook output. That is indeed meaningless. Perhaps, we could convert this notebook into a .py script so that it would be stored in the repo, but without output?

[AgnieszkaZaba]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.
Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

Ok, you're right. I'm still waiting for Jupyter to finish running :)

Ah, I didn't realise you are referring to running the GPU tests locally on a machine without GPU to generate notebook output. That is indeed meaningless. Perhaps, we could convert this notebook into a .py script so that it would be stored in the repo, but without output?

Hmm I feel like this output is important and should be stored somewhere. Not necessarily in notebook. I'm trying to run it on Colab. Keep disconnecting me but program is still running and giving results, so maybe will be ok.

[slayoo]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.
Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

Ok, you're right. I'm still waiting for Jupyter to finish running :)

Ah, I didn't realise you are referring to running the GPU tests locally on a machine without GPU to generate notebook output. That is indeed meaningless. Perhaps, we could convert this notebook into a .py script so that it would be stored in the repo, but without output?

Hmm I feel like this output is important and should be stored somewhere. Not necessarily in notebook. I'm trying to run it on Colab. Keep disconnecting me but program is still running and giving results, so maybe will be ok.

The output is also persistently stored in the UJ AP repo (pages 29-30): https://www.ap.uj.edu.pl/diplomas/attachments/file/download/168064/

[AgnieszkaZaba]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.
Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

Ok, you're right. I'm still waiting for Jupyter to finish running :)

Ah, I didn't realise you are referring to running the GPU tests locally on a machine without GPU to generate notebook output. That is indeed meaningless. Perhaps, we could convert this notebook into a .py script so that it would be stored in the repo, but without output?

Hmm I feel like this output is important and should be stored somewhere. Not necessarily in notebook. I'm trying to run it on Colab. Keep disconnecting me but program is still running and giving results, so maybe will be ok.

The output is also persistently stored in the UJ AP repo (pages 29-30): https://www.ap.uj.edu.pl/diplomas/attachments/file/download/168064/

That is great! I will add this link to the notebook.
And maybe .py will be better. Not sure. Because even with GPU run time is long. So maybe shorten in notebook? Or change to .py as you suggested..

[slayoo]

@slayoo only one notebook should be incorrect: it's performance comparison Inside is a comparison between using numba and ThrustRC. And running it without numba is not only very long but also meaningless.
Don't know why the others are wrong, I change them many times. Will try once again.

@AgnieszkaZaba, Numba is enabled when running the notebooks on CI. There are no GPUs available on CI, but we have FakeThrustRTC mock class specifically developed to emulate ThrustRTC using Numba. I'd say running these is not meaningless as it ensures that the notebook code is compatible with the evolving PySDM API, so if one runs it on a local machine, it's likely that it will actually work.

Ok, you're right. I'm still waiting for Jupyter to finish running :)

Ah, I didn't realise you are referring to running the GPU tests locally on a machine without GPU to generate notebook output. That is indeed meaningless. Perhaps, we could convert this notebook into a .py script so that it would be stored in the repo, but without output?

Hmm I feel like this output is important and should be stored somewhere. Not necessarily in notebook. I'm trying to run it on Colab. Keep disconnecting me but program is still running and giving results, so maybe will be ok.

The output is also persistently stored in the UJ AP repo (pages 29-30): https://www.ap.uj.edu.pl/diplomas/attachments/file/download/168064/

That is great! I will add this link to the notebook. And maybe .py will be better. Not sure. Because even with GPU run time is long. So maybe shorten in notebook? Or change to .py as you suggested..

Thank you!
GPU times should be shorter than CPU=Numba (for large-enough domain, as depicted in the thesis plots).

[AgnieszkaZaba]

@slayoo thanks for rebasing, I didn't succeed with that.

Right now I'm trying to run notebooks but after 5 hours still do not have results

[slayoo]

@slayoo thanks for rebasing, I didn't succeed with that.

This was a merge. For PR branches, merges are usually straightforward, and thus much simpler than rebasing. Rebasing works best for the always-clean main branch.

Right now I'm trying to run notebooks but after 5 hours still do not have results

Which notebooks take so long?

[AgnieszkaZaba]

@slayoo thanks for rebasing, I didn't succeed with that.

This was a merge. For PR branches, merges are usually straightforward, and thus much simpler than rebasing. Rebasing works best for the always-clean main branch.

aa ok, still dont fully understand, but I will get there.

Right now I'm trying to run notebooks but after 5 hours still do not have results

Which notebooks take so long?

From deJong_Mackay_et_al_2023 fig9 and figs_3_4_5

[slayoo]

@slayoo thanks for rebasing, I didn't succeed with that.

This was a merge. For PR branches, merges are usually straightforward, and thus much simpler than rebasing. Rebasing works best for the always-clean main branch.

aa ok, still dont fully understand, but I will get there.

https://learngitbranching.js.org/

[AgnieszkaZaba]

@slayoo thanks for rebasing, I didn't succeed with that.

This was a merge. For PR branches, merges are usually straightforward, and thus much simpler than rebasing. Rebasing works best for the always-clean main branch.

aa ok, still dont fully understand, but I will get there.

Right now I'm trying to run notebooks but after 5 hours still do not have results

Which notebooks take so long?

From deJong_Mackay_et_al_2023 fig9 and figs_3_4_5

For both notebook it was around 11 hours 😮

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.22%. Comparing base (5fbc411) to head (2fb371e).
Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1453   +/-   ##
=======================================
  Coverage   85.22%   85.22%           
=======================================
  Files         376      376           
  Lines        9266     9266           
=======================================
  Hits         7897     7897           
  Misses       1369     1369           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.