Handle latest PyMC/PyTensor breaking changes

[ricardoV94]

No description provided.

[ricardoV94]

@ciguaran @aloctavodia, there's an SMC error that I think is unrelated, probably something changed in blackjax?

AttributeError: module 'blackjax.smc.adaptive_tempered' has no attribute 'adaptive_tempered_smc'

How should I fix it?
Also SMC is inside the inference module, shouldn't it be in its own module. In the PyMC repo inference is specific to VI stuff?

[ricardoV94]

Still some broken stuff, but almost done

[ricardoV94]

Ok 4 failing tests, one marginal, two smc, and one statespace.

@jessegrabowski: FAILED pymc_experimental/tests/statespace/test_VARMAX.py::test_impulse_response[scalar_shock_size] - ValueError: initial_shock has 0 dims but 1 dim labels were provided.

Easy fix?

[jessegrabowski]

Should be fixed by #346

[ricardoV94]

@jessegrabowski can we fix just the dim here? Otherwise this PR depends on that and that one will depend on this, unless you put all commits from this PR there?

[jessegrabowski]

Sure np

[ricardoV94]

Sure np

Can you push or tell me what to change. If it's convoluted I can also just mark it as xfail?

[jessegrabowski]

I'm doing it now, give me about an hour

[ricardoV94]

I'm doing it now, give me about an hour

Counting

[zaxtax]

What other files are still giving errors?

[ricardoV94]

One marginal test and two SMC tests

[ciguaran]

@ricardoV94 hei! could you share the version of blackjax you are using? most likely you can fix it by replacing 'blackjax.smc.adaptive_tempered' with blackjax.adaptive_tempered_smc

[ricardoV94]

@ciguaran other stuff is failing downstream after I update the function path. I'll mark it as xfail for now

[ricardoV94]

To unblock parallel work I marked two tests as xfail and opened issues for it

• Recover nested marginals test is failing #355
• Blacjax SMC test is failing #354

[jessegrabowski]

Is this ready to merge?

[zaxtax]

We should bump the middle version number as part of this change