Bump PyTensor dependency #6531
Bump PyTensor dependency #6531
Conversation
|
isn't this also in the precommit? |
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #6531 +/- ##
==========================================
+ Coverage 91.91% 93.02% +1.11%
==========================================
Files 89 147 +58
Lines 14936 27932 +12996
==========================================
+ Hits 13729 25985 +12256
- Misses 1207 1947 +740
|
|
|
Yeah, I'm investigating it. It has a C implementation... |
I don't think it is anymore |
|
The external samplers failure is what I feared would happen with the nutpie circular dependency? |
|
The warning should be fixed by pymc-devs/pytensor#225 |
ricardoV94
force-pushed
the
bump_pytensor_dep
branch
2 times, most recently
from
37ba5a4
to
b24d458
Compare
|
Lots of install conflicts :(. |
|
It's because nutpie pins the old pymc's pytensor dependency on conda. We can't test it directly as I suspected. Will remove the nutpie tests for now |
Was fixed in pymc-devs/pytensor#205
Conda nutpie, and soon pip, depend on the last release of PyMC and by association PyTensor, leading to dependency issues when trying to test compatibility after PyMC bumps the PyTensor dependency.
ricardoV94
force-pushed
the
bump_pytensor_dep
branch
from
b24d458
to
f59adee
Compare
| if nuts_sampler != "pymc": | ||
| pytest.importorskip(nuts_sampler) |
There was a problem hiding this comment.
I checked this works as expected and skips only the parametrization that fails to import, not all of them afterwards. Can be seen in the CI as well: https://github.com/pymc-devs/pymc/actions/runs/4282342784/jobs/7456497043#step:8:79
No description provided.