Add warning if a Minibatched variable is used without total_size

[zaxtax]

Description

This should help catch a well-known Minibatch footgun where a user forgets to include the total_size argument

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and/or example notebooks)
• If you are a pro: each commit corresponds to a relevant logical change

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):

---

📚 Documentation preview 📚: https://pymc--7742.org.readthedocs.build/en/7742/

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 92.82%. Comparing base (fb4b201) to head (e10511a).
Report is 4 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #7742   +/-   ##
=======================================
  Coverage   92.82%   92.82%           
=======================================
  Files         107      107           
  Lines       18324    18326    +2     
=======================================
+ Hits        17010    17012    +2     
  Misses       1314     1314           

Files with missing lines	Coverage Δ
pymc/model/core.py	92.99% <100.00%> (+0.01%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jessegrabowski]

LGTM. My only suggestion is that we have some way for a user who knows he doesn't want to set total_size to somehow avoid the warning. I can't think of a case where anyone would want to do this, but something like total_size=False to silence the warning might be nice?

Might also be too convoluted.

[zaxtax]

I don't think you ever want to not set total_size if you are doing minibatching. If you are doing something fancy you probably are just setting it to something other than len(data)

…

On Tue, 1 Apr 2025, 10:10 Jesse Grabowski, ***@***.***> wrote: ***@***.**** approved this pull request. LGTM. My only suggestion is that we have some way for a user who knows he doesn't want to set total_size to somehow avoid the warning. I can't think of a case where anyone would want to do this, but something like total_size=False to silence the warning might be nice? Might also be too convoluted. — Reply to this email directly, view it on GitHub <#7742 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAACCUIZBJRM5CAGC5PTEQL2XJCWHAVCNFSM6AAAAAB2CSI7Q6VHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZDOMZRHA2TGMBXGI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[ricardoV94]

this should have checked if minibatch is anywhere in the ancestors of value not just immediately

[zaxtax]

Do we need to do that? This check is only for observed random variables. I thought for those we don't allow the value to depend on another node