Unpin scipy upper version

[ferrine]

Description

Related Issue

• Closes Scipy 1.14.0 changed the upcasting rule of several special functions when passing integers #862
• Closes Simplify elemwise? #973
• Related to Several scipy special functions now upcast integers to float64 #859

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):

[ferrine]

besides expected errors there is one that comes from mypy

!!!!!!!!!
2 files unexpectedly failed.
/home/runner/work/pytensor/pytensor/pytensor/gradient.py
/home/runner/work/pytensor/pytensor/pytensor/graph/basic.py
These files did not fail before, so please check the above output for errors in {PosixPath('/home/runner/work/pytensor/pytensor/pytensor/graph/basic.py'), PosixPath('/home/runner/work/pytensor/pytensor/pytensor/gradient.py')} and fix them.
You can run `python scripts/run_mypy.py --verbose` to reproduce this test locally.

[ferrine]

I also think of simplifying this part of code, it has a lot of redundant logic that is already managed by simple np.asarray

diff --git a/pytensor/tensor/elemwise.py b/pytensor/tensor/elemwise.py
index de966f1a7..a0892dfd0 100644
--- a/pytensor/tensor/elemwise.py
+++ b/pytensor/tensor/elemwise.py
@@ -767,31 +767,17 @@ class Elemwise(OpenMPOp):
         for i, (variable, storage, nout) in enumerate(
             zip(variables, output_storage, node.outputs)
         ):
-            if getattr(variable, "dtype", "") == "object":
-                # Since numpy 1.6, function created with numpy.frompyfunc
-                # always return an ndarray with dtype object
-                variable = np.asarray(variable, dtype=nout.dtype)
+
+            variable = np.asarray(variable, dtype=nout.dtype)

             if i in self.inplace_pattern:
                 odat = inputs[self.inplace_pattern[i]]
                 odat[...] = variable
                 storage[0] = odat

-            # Sometimes NumPy return a Python type.
-            # Some PyTensor op return a different dtype like floor, ceil,
-            # trunc, eq, ...
-            elif not isinstance(variable, np.ndarray) or variable.dtype != nout.dtype:
-                variable = np.asarray(variable, nout.dtype)
-                # The next line is needed for numpy 1.9. Otherwise
-                # there are tests that fail in DebugMode.
-                # Normally we would call pytensor.misc._asarray, but it
-                # is faster to inline the code. We know that the dtype
-                # are the same string, just different typenum.
-                if np.dtype(nout.dtype).num != variable.dtype.num:
-                    variable = variable.view(dtype=nout.dtype)
-                storage[0] = variable
+            storage[0] = variable
             # numpy.real return a view!
-            elif not variable.flags.owndata:
+            if not variable.flags.owndata:
                 storage[0] = variable.copy()
-            else:
-                storage[0] = variable

[ferrine]

@ricardoV94 how are these tests are supposed to pass?

pytensor/tests/scalar/test_loop.py

Line 198 in 4d0103b

def test_inner_composite(mode):

Why did they pass before?

[ferrine]

Wow, looking into the broken test closer, it should have been a failing test!

before the PR

after the change the dtype of out is float16 as declared in the graph and we see overflow

[codecov]

Codecov Report

Attention: Patch coverage is 75.00000% with 5 lines in your changes missing coverage. Please review.

Project coverage is 81.69%. Comparing base (29183c7) to head (d82b1e5).
Report is 101 commits behind head on main.

Files with missing lines	Patch %	Lines
pytensor/scalar/basic.py	75.00%	3 Missing ⚠️
pytensor/gradient.py	60.00%	0 Missing and 2 partials ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #972   +/-   ##
=======================================
  Coverage   81.69%   81.69%           
=======================================
  Files         182      182           
  Lines       47585    47591    +6     
  Branches    11584    11586    +2     
=======================================
+ Hits        38875    38880    +5     
  Misses       6518     6518           
- Partials     2192     2193    +1     

Files with missing lines	Coverage Δ
pytensor/graph/basic.py	88.66% <100.00%> (ø)
pytensor/tensor/elemwise.py	88.54% <100.00%> (+0.16%)	⬆️
pytensor/gradient.py	77.49% <60.00%> (-0.15%)	⬇️
pytensor/scalar/basic.py	80.41% <75.00%> (+0.03%)	⬆️

[ricardoV94]

Thanks @ferrine

[ricardoV94]

What's with mypy? I guess it's not related, but can you fix it?

[ricardoV94]

@ferrine can you clean the commits into just 2:

1. scipy stuff
2. mypy stuff
   ?

[ferrine]

Yes, will do soon