Drop Python 3.10 and NumPy < 2 #1253

Armavica · 2025-02-27T15:58:53Z

Description

Related Issue

Closes #
Related to #

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

📚 Documentation preview 📚: https://pytensor--1253.org.readthedocs.build/en/1253/

codecov · 2025-02-27T16:33:16Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.65%. Comparing base (b05acfd) to head (04b2d67).
⚠️ Report is 16 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1253      +/-   ##
==========================================
- Coverage   81.67%   81.65%   -0.03%     
==========================================
  Files         232      232              
  Lines       53132    53078      -54     
  Branches     9410     9405       -5     
==========================================
- Hits        43396    43340      -56     
- Misses       7283     7285       +2     
  Partials     2453     2453

Files with missing lines	Coverage Δ
pytensor/link/c/basic.py	`87.74% <100.00%> (-0.02%)`	⬇️
pytensor/link/numba/dispatch/elemwise.py	`94.82% <100.00%> (ø)`
pytensor/npy_2_compat.py	`100.00% <100.00%> (ø)`
pytensor/tensor/basic.py	`91.80% <100.00%> (ø)`
pytensor/tensor/blas.py	`73.52% <100.00%> (ø)`
pytensor/tensor/einsum.py	`97.11% <100.00%> (+0.01%)`	⬆️
pytensor/tensor/elemwise.py	`90.01% <100.00%> (ø)`
pytensor/tensor/extra_ops.py	`88.54% <100.00%> (-0.02%)`	⬇️
pytensor/tensor/math.py	`92.86% <100.00%> (-0.02%)`	⬇️
pytensor/tensor/nlinalg.py	`94.98% <100.00%> (ø)`
... and 5 more

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

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

ricardoV94 · 2025-09-29T14:55:47Z

Does this mean we can also drop numpy<2 and related CI matrix?

Armavica · 2025-09-29T14:57:16Z

Yeah I think that the plan was to drop numpy < 2.0 and python 3.10 together if I remember correctly

ricardoV94 · 2025-09-29T15:02:25Z

Let's do it. There is stuff in npy_compat as well

scripts/slowest_tests/update-slowest-times-issue.sh

Armavica · 2025-09-29T16:15:39Z

There is stuff remaining but I want to check this for now.

.github/workflows/test.yml

pytensor/tensor/subtensor.py

.github/workflows/test.yml

ricardoV94 · 2025-09-30T09:56:41Z

pytensor/tensor/math.py

        else:
            # The value here doesn't matter, it won't be used
-            c_axis = numpy_axis_is_none_flag
+            c_axis = np.iinfo(np.int32).min  # the value of NPY_RAVEL_AXIS


Seems better to have this computed in one of the util files with the more readable name?

But in this case, if the value actually doesn't matter as the comment says, perhaps we can even use 0?

Yeah that seems right

Armavica · 2025-09-30T10:07:36Z

I was wondering if I need to increment the c code version after removing the compatibility headers?

ricardoV94 · 2025-09-30T10:18:49Z

I was wondering if I need to increment the c code version after removing the compatibility headers?

I don't think so, the generated C code didn't change if you are using numpy>2.0, and you can't use the old one anymore now. Could be wrong ofc

Armavica · 2025-09-30T11:55:55Z

I don't think so, the generated C code didn't change if you are using numpy>2.0

Even with this commit? a0ad246

ricardoV94 · 2025-09-30T12:02:40Z

Even with this commit? a0ad246

Can you check if we bumped the C versions when we added that header thing to those Ops? If we did, then maybe let's be safe and bump again

ricardoV94 · 2025-09-30T12:08:02Z

.github/workflows/test.yml

+          - fast-compile: 1
+          - float32: 1
+          - python-version: "3.13"


I couldn't comment with suggestion on the phone, but these are excluding all fast-compile/float32/python 3.13 runs. They only made sense in combination with numpy version (we were trying not to blow the CI when we added the numpy variants)

Suggested change

- fast-compile: 1

- float32: 1

- python-version: "3.13"

Armavica · 2025-09-30T14:19:10Z

Can you check if we bumped the C versions when we added that header thing to those Ops? If we did, then maybe let's be safe and bump again

This is the commit that introduced that header, it changed the C code in other ways too but it did bump the C versions: 0aa10c0

ricardoV94 · 2025-09-30T14:38:21Z

Right, let's rebump then. Thanks for checking it

ricardoV94 · 2025-10-03T20:33:48Z

pytensor/link/c/basic.py


        # We must always add the numpy ABI version here as
        # DynamicModule always add the include <numpy/arrayobject.h>
+        ndarray_c_version = _get_ndarray_c_version()


How slow is this? Do we want to execute ot only once perhaps?

ricardoV94 · 2025-10-03T20:35:26Z

looks good just a small question

ricardoV94 · 2025-10-05T14:02:58Z

Thanks a ton @Armavica

Anything else missing?

Armavica · 2025-10-05T14:51:35Z

Thanks a ton @Armavica

Anything else missing?

I can't think of anything else

ricardoV94 · 2025-10-06T13:55:19Z

The PyPI CI seems to be failing

Armavica · 2025-10-07T08:57:12Z

Not sure what this is about, I will debug in a few days

ricardoV94 · 2025-10-07T10:36:55Z

I think it's trying to build something with python 3.10. Also are you rebased from main?

ricardoV94 · 2025-10-07T10:40:10Z

Specifically this job may need the python version specified (the one above build_universal_wheel specifies it for example):

pytensor/.github/workflows/pypi.yml

Line 186 in 4312d8c

name: Check dist

Armavica · 2025-10-07T11:14:16Z

Specifically this job may need the python version specified (the one above build_universal_wheel specifies it for example):

pytensor/.github/workflows/pypi.yml

Line 186 in 4312d8c

name: Check dist

Thank you, it looks like that did the trick

ricardoV94 · 2025-10-07T11:31:31Z

LGTM. I asked @lucianopaz and @maresb to also take a look in case they spot something else problematic. Specifically I suspect @maresb needs to be in the loop for the conda release anyway

maresb

Thanks @Armavica! I'll be ready for this on the conda-forge side.

ricardoV94 · 2025-10-07T22:33:10Z

Thanks again @Armavica

Armavica force-pushed the drop-3.10 branch from 0b80ced to 64a8c36 Compare September 29, 2025 14:53

ricardoV94 reviewed Sep 29, 2025

View reviewed changes

scripts/slowest_tests/update-slowest-times-issue.sh Show resolved Hide resolved

Armavica force-pushed the drop-3.10 branch from 64a8c36 to d6370a3 Compare September 29, 2025 16:14

Armavica force-pushed the drop-3.10 branch from d6370a3 to 2d9b73c Compare September 29, 2025 16:31

ricardoV94 reviewed Sep 29, 2025

View reviewed changes

.github/workflows/test.yml Outdated Show resolved Hide resolved

ricardoV94 reviewed Sep 29, 2025

View reviewed changes

pytensor/tensor/subtensor.py Outdated Show resolved Hide resolved

Armavica force-pushed the drop-3.10 branch from 2d9b73c to 79e68e6 Compare September 30, 2025 09:27

Armavica changed the title ~~Drop Python 3.10~~ Drop Python 3.10 and NumPy < 2 Sep 30, 2025

ricardoV94 reviewed Sep 30, 2025

View reviewed changes

Armavica force-pushed the drop-3.10 branch 3 times, most recently from 463e252 to 5a8976b Compare October 1, 2025 11:56

Armavica marked this pull request as ready for review October 1, 2025 12:55

Armavica requested a review from ricardoV94 October 3, 2025 17:57

ricardoV94 reviewed Oct 3, 2025

View reviewed changes

Armavica force-pushed the drop-3.10 branch 2 times, most recently from 8b375f4 to e28ba52 Compare October 5, 2025 09:03

ricardoV94 approved these changes Oct 5, 2025

View reviewed changes

ricardoV94 added major release labels Oct 5, 2025

Armavica added 11 commits October 5, 2025 16:26

Drop Python 3.10

615cfcb

Remove npy<2 compatibility for normalize_axis_{index,tuple}

38bf02c

Remove npy<2 compatibility for OverflowError

71a23d9

Remove npy<2 compatibility for einsum helpers

6d9906b

Remove npy<2 compatibility for numpy_maxdims

3ef22b3

Remove npy<2 compatibility for NPY_RAVEL_AXIS value

74e8afe

Remove npy<2 compatibility for ndarray_c_version

0cf546d

Remove using_numpy_2

72b5dfc

Remove numpy-version in the CI matrix

a3fa078

Remove the numpy < 2 compatibility header

382ee3f

Remove numpy < 2.0

b6e1b77

Armavica force-pushed the drop-3.10 branch from e28ba52 to b6e1b77 Compare October 5, 2025 14:29

Specify the python version for the check_dist CI job

04b2d67

ricardoV94 requested review from lucianopaz and maresb October 7, 2025 11:30

maresb approved these changes Oct 7, 2025

View reviewed changes

ricardoV94 merged commit 5227759 into pymc-devs:main Oct 7, 2025
59 checks passed

Armavica deleted the drop-3.10 branch October 8, 2025 12:36

Drop Python 3.10 and NumPy < 2 #1253

Drop Python 3.10 and NumPy < 2 #1253

Uh oh!

Conversation

Armavica commented Feb 27, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Related Issue

Checklist

Type of change

Uh oh!

codecov bot commented Feb 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

ricardoV94 commented Sep 29, 2025

Uh oh!

Armavica commented Sep 29, 2025

Uh oh!

ricardoV94 commented Sep 29, 2025

Uh oh!

Uh oh!

Armavica commented Sep 29, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ricardoV94 Sep 30, 2025

Choose a reason for hiding this comment

Uh oh!

Armavica Sep 30, 2025

Choose a reason for hiding this comment

Uh oh!

ricardoV94 Sep 30, 2025

Choose a reason for hiding this comment

Uh oh!

Armavica commented Sep 30, 2025

Uh oh!

ricardoV94 commented Sep 30, 2025

Uh oh!

Armavica commented Sep 30, 2025

Uh oh!

ricardoV94 commented Sep 30, 2025

Uh oh!

ricardoV94 Sep 30, 2025

Choose a reason for hiding this comment

Uh oh!

Armavica commented Sep 30, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ricardoV94 commented Sep 30, 2025

Uh oh!

ricardoV94 Oct 3, 2025

Choose a reason for hiding this comment

Uh oh!

ricardoV94 commented Oct 3, 2025

Uh oh!

ricardoV94 commented Oct 5, 2025

Uh oh!

Armavica commented Oct 5, 2025

Uh oh!

ricardoV94 commented Oct 6, 2025

Uh oh!

Armavica commented Oct 7, 2025

Uh oh!

ricardoV94 commented Oct 7, 2025

Uh oh!

ricardoV94 commented Oct 7, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Armavica commented Oct 7, 2025

Uh oh!

ricardoV94 commented Oct 7, 2025

Uh oh!

maresb left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

ricardoV94 commented Oct 7, 2025

Uh oh!

Armavica commented Feb 27, 2025 •

edited by github-actions bot

Loading

codecov bot commented Feb 27, 2025 •

edited

Loading

Armavica commented Sep 30, 2025 •

edited

Loading

ricardoV94 commented Oct 7, 2025 •

edited

Loading