Bump PyTensor dependency and support Python 3.14

[ricardoV94]

Closes #7934

[codecov]

Codecov Report

❌ Patch coverage is 85.71429% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 91.45%. Comparing base (cadb97a) to head (1a549c9).
⚠️ Report is 5 commits behind head on main.

Files with missing lines	Patch %	Lines
pymc/distributions/simulator.py	0.00%	2 Missing ⚠️
pymc/variational/approximations.py	0.00%	1 Missing ⚠️
pymc/variational/opvi.py	50.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8010      +/-   ##
==========================================
+ Coverage   90.22%   91.45%   +1.22%     
==========================================
  Files         116      116              
  Lines       18972    18971       -1     
==========================================
+ Hits        17117    17349     +232     
+ Misses       1855     1622     -233     

Files with missing lines	Coverage Δ
pymc/distributions/multivariate.py	93.92% <100.00%> (+0.01%)	⬆️
pymc/distributions/shape_utils.py	91.87% <100.00%> (ø)
pymc/distributions/timeseries.py	94.57% <100.00%> (-0.05%)	⬇️
pymc/logprob/transforms.py	95.43% <100.00%> (ø)
pymc/pytensorf.py	88.11% <100.00%> (ø)
pymc/smc/sampling.py	77.08% <100.00%> (+1.38%)	⬆️
pymc/variational/operators.py	93.47% <100.00%> (ø)
pymc/variational/approximations.py	90.86% <0.00%> (ø)
pymc/variational/opvi.py	86.75% <50.00%> (ø)
pymc/distributions/simulator.py	84.28% <0.00%> (ø)

... and 8 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]

I couldn't reproduce the issue locally. CI is running on 3.14. Installing the same numba/scipy versions on my local 3.12 wasn't enough to replicate. I wouldn't block the dependency bump on this as it's rather niche functionality.

Tracking issue: #8012

[ricardoV94]

RTD is failing with a strange conda error?

https://app.readthedocs.org/projects/pymc/builds/30757275/

error    Error opening "/home/docs/.asdf/installs/python/mambaforge-4.10.3-10/pkgs/pytensor-2.36.0-py314h8fa1eac_1.conda": No such file or directory
error    File not valid: SHA256 sum doesn't match expectation "/home/docs/.asdf/installs/python/mambaforge-4.10.3-10/pkgs/pytensor-2.36.0-py314h8fa1eac_1.conda"
    Expected: 0c2f7154116702b630c239df3780ec9963c8de36876f6cb31e7632a77a80dd53
    Actual: e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
    
# >>>>>>>>>>>>>>>>>>>>>> ERROR REPORT <<<<<<<<<<<<<<<<<<<<<<
    Traceback (most recent call last):
      File "/home/docs/.asdf/installs/python/mambaforge-4.10.3-10/lib/python3.9/site-packages/conda/exceptions.py", line 1079, in __call__
        return func(*args, **kwargs)
      File "/home/docs/.asdf/installs/python/mambaforge-4.10.3-10/lib/python3.9/site-packages/conda_env/cli/main.py", line 80, in do_call
        exit_code = getattr(module, func_name)(args, parser)
      File "/home/docs/.asdf/installs/python/mambaforge-4.10.3-10/lib/python3.9/site-packages/conda_env/cli/main_create.py", line 141, in execute
        result[installer_type] = installer.install(prefix, pkg_specs, args, env)
      File "/home/docs/.asdf/installs/python/mambaforge-4.10.3-10/lib/python3.9/site-packages/mamba/mamba_env.py", line 125, in mamba_install
        downloaded = transaction.prompt(repos)
    RuntimeError: Found incorrect download: pytensor. Aborting
`$ /home/docs/.asdf/installs/python/mambaforge-4.10.3-10/bin/mamba create --quiet --name 8010 --file conda-envs/environment-docs.yml`

CC @maresb @OriolAbril

Maybe there's some cache in the CI that needs to be erased?

[ricardoV94]

Bug is pymc-devs/pytensor#1796

Hopefully we can fix in the next release and update the remaining tests to the new API

[OriolAbril]

RTD is failing with a strange conda error?

I triggered a re-run which seems to be working and already created the env successfully, no idea what could have caused that particular issue though. AFAIK there is no caching on RTD and even if there were, the previous builds used the same exact conda build

[ricardoV94]

Thanks @OriolAbril