Allow Prior to create pymc.dims variables #643
Conversation
ricardoV94
force-pushed
the
xprior
branch
3 times, most recently
from
0f03f4e to
5dc08d7
Compare
ricardoV94
force-pushed
the
xprior
branch
2 times, most recently
from
0714a3a to
5284433
Compare
| if isinstance(value.type, pt.TensorType): | ||
| value = value.eval() | ||
|
|
||
| # Avoid XTensor import warnings, remove this when the warnings are gone |
williambdean
left a comment
williambdean
left a comment
There was a problem hiding this comment.
Overall looks pretty good.
| if xdist: | ||
| import pytensor.xtensor as ptx | ||
|
|
||
| for module in (ptx.math, ptx.linalg, ptx.signal, ptx): |
There was a problem hiding this comment.
This seems to be causing some errors in the CI/CD
There was a problem hiding this comment.
signal is not released yet I guess
| prior = dist.sample_prior(coords=coords) | ||
| """ | ||
| # TODO: Why do we need a sample_prior function? |
There was a problem hiding this comment.
sample_prior was to allow for VariableFactory instances to have sample_prior
functionality. User doesn't need to define a method.
There was a problem hiding this comment.
Oh I didn't mean to leave the comment, sorry
|
|
||
| def test_custom_transform() -> None: | ||
| new_transform_name = "foo_bar" | ||
| dist = Prior("Normal", transform=new_transform_name) |
There was a problem hiding this comment.
So this raises later now?
There was a problem hiding this comment.
We could raise for tensor, it's unlikely it doesn't exist for that but it exists for xtensor. We also check for pymc distribution even though it may not be used at all
3 participants
Add an
xdistboolean flag to createpymc.dimsvariables instead.This is a breaking change, existing custom classes will need to accept
xdistflag increate_variable.Local/odd imports are inplace until we remove experimental warnings from importing pymc.dims or pytensor.xtensor