Update Aesara version and unpin numpy

[twiecki]

Closes #5321.
Reverts temporary #5332.

[ricardoV94]

Any idea what's wrong with the conda setup?

[twiecki]

I think I was just too fast before conda-forge had aesara packages up.

[codecov]

Codecov Report

Merging #5369 (99dafaf) into main (d52655d) will increase coverage by 2.34%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #5369      +/-   ##
==========================================
+ Coverage   80.36%   82.71%   +2.34%     
==========================================
  Files          89      132      +43     
  Lines       14808    26042   +11234     
==========================================
+ Hits        11901    21541    +9640     
- Misses       2907     4501    +1594     

Impacted Files	Coverage Δ
pymc/distributions/multivariate.py	75.57% <100.00%> (+1.44%)	⬆️
pymc/ode/ode.py	84.84% <100.00%> (ø)
pymc/tests/test_smc.py	97.67% <100.00%> (ø)
pymc/sampling_jax.py	0.00% <0.00%> (-97.46%)	⬇️
pymc/parallel_sampling.py	86.71% <0.00%> (-1.00%)	⬇️
pymc/distributions/continuous.py	96.93% <0.00%> (-0.50%)	⬇️
pymc/distributions/__init__.py	100.00% <0.00%> (ø)
pymc/distributions/distribution.py	91.40% <0.00%> (ø)
pymc/tests/test_transforms.py	94.18% <0.00%> (ø)
pymc/tests/test_minibatches.py	98.50% <0.00%> (ø)
... and 42 more

[twiecki]

pymc/tests/test_ode.py::TestDiffEqModel::test_op_equality PASSED         [ 44%]
896
/home/runner/work/_temp/25769632-011a-4256-95b2-edae057d3830.sh: line 2:  7587 Killed                  python -m pytest -vv --cov=pymc --cov-append --cov-report=xml --cov-report term --durations=50 $TEST_SUBSET
897
pymc/tests/test_ode.py::TestDiffEqModel::test_scalar_ode_1_param 

[twiecki]

Not sure why this stalls out, it passes locally. Any idea @ricardoV94 @michaelosthege ?

[ricardoV94]

Not sure why this stalls out, it passes locally. Any idea @ricardoV94 @michaelosthege ?

Are you testing with the latest Aesara? It fails locally for me. I think I know where the problem is...

[twiecki]

Alright, I wasn't. What's the problem?

[ricardoV94]

Alright, I wasn't. What's the problem?

Still trying to pin it down. Those ODE Ops are super hackish xD

[ricardoV94]

I think there is a broken cyclical optimization on the Aesara side

[ricardoV94]

Yup, found it.