DOC: JNB-RTD consistency for idars.ipynb

[DavidBAEpstein]

• MAINT: catch up with another branch
• BUG: fix on Macos without GPU by WORKERS=0
• DOC; change comment about ON_GPU
• DOC: Run on Colab with GPU and ON_GPU--True. Save with output.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[codecov]

Codecov Report

Merging #421 (f4350c7) into develop (fcfcfc2) will not change coverage.
The diff coverage is n/a.

❗ Current head f4350c7 differs from pull request most recent head 9af672a. Consider uploading reports for the commit 9af672a to get more accurate results

@@           Coverage Diff            @@
##           develop     #421   +/-   ##
========================================
  Coverage    98.63%   98.63%           
========================================
  Files           60       60           
  Lines         5916     5916           
  Branches      1053     1053           
========================================
  Hits          5835     5835           
  Misses          69       69           
  Partials        12       12           

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

mostafajahanifar commented on 2022-08-05T08:32:49Z
----------------------------------------------------------------

Thanks David for the correction. If this fix is needed for MacOs, I would suggest doing the same thing for all notebooks.

DavidBAEpstein commented on 2022-08-05T09:13:09Z
----------------------------------------------------------------

My "fix" above with defining WORKERS does not succeed with 07-advanced-modeling.ipynb.. I get an error saying that Cuda-enabled GPU is needed. Much the same error message is obtained when running 07-advanced-modeling.ipynb.on Colab without GPU.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

mostafajahanifar commented on 2022-08-05T08:32:49Z
----------------------------------------------------------------

the legend is not correct, showing red for both tumour and non-tumour. Maybe @srijay can help with this?

[mostafajahanifar]

Thanks David for the upgrading this notebook.

The only minor comment from me that requires attention is fixing the legend for the tumour prediction figure, where both tumour and non-tumour have the same colour.

[DavidBAEpstein]

I might be able to guess how to fix this, and will have a try

---

View entire conversation on ReviewNB

[DavidBAEpstein]

My "fix" above with defining WORKERS does not succeed with 07-advanced-modeling.ipynb.. I get an error saying that Cuda-enabled GPU is needed. I don't understand what is going on, but I think the code requires a version of torch that requires GPU.

So I am on the point of removing all mention of ON_GPU from the 07 notebook. The fix does work on this notebook idars.ipynb

---

View entire conversation on ReviewNB

[DavidBAEpstein]

I changed non-tumour colour to green. The overlay image changed dramatically, in the way I expected. The legend

remained red throughout. My guess is that this is related to the warning message immediately above the image, namely

|2022-07-20|15:37:40.408| [WARNING] ....

Perhaps overlay_prediction_mask cannot cope with scale > 1, and the scale for the legend is incorrect.

The code in the develop version is identical with the code here, and it gives a correct legend. So I think the behaviour is not properly controlled by overlay_prediction_mask, so that the same code gives very different results under very very slightly different conditions.

---

View entire conversation on ReviewNB

[DavidBAEpstein]

inference-pipelines/idars.ipynb runs successfully under Macos without GPU (ON_GPU == False)
In order for it to run correctly under Colab with GPU (ON_GPU == True) it was necessary to replace "pip install tiatoolbox" with
"pip install git+https://github.com/TissueImageAnalytics/tiatoolbox.git@develop" in order to use @Srijay-lab's fix of overlay-prediction-mask. This will need to be changed back at the next release of tiatoolbox.

[DavidBAEpstein]

Thanks David for the upgrading this notebook.

The only minor comment from me that requires attention is fixing the legend for the tumour prediction figure, where both tumour and non-tumour have the same colour.

The error was due to a bug in the function overlay_prediction_mask, which is fixed in the latest version of develop.

[DavidBAEpstein]

@mostafajahanifar @shaneahmed This PR is now ready for review. It has had a chequered career because of the bug in the function overlay-prediction-mask.

[shaneahmed]

Thanks @DavidBAEpstein for your work on this.
Looks good!
I have added hidden cells as they create redundant text in the documentation. The sections and subsections look fine to me after removing these cells from readthedocs.