-
Notifications
You must be signed in to change notification settings - Fork 135
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update HERON Submodule #2248
Update HERON Submodule #2248
Conversation
@joshua-cogliati-inl @j-bryan I'm stumped as to why the RAVEN tests are failing on a PR that only has a submodule change. Is there a possibility the failing tests have to do with the recent Distribution changes? |
I am wondering if there is a problem with a package. As in here is code that causes this error:
and I noticed that we went from liblapack 3.9.0 20_linux64_openblas to liblapack 3.9.0 21_linux64_openblas |
@wangcj05 this looks like an issue with the PCA-reduced multivariate normal not having a symmetric positive semi-definite input matrix. This seems to be a recent failure due to a library change. Did you have to wrestle with this SPD matrix input while setting up this problem? |
@PaulTalbot-INL There's a try/except block at that part of |
Ah, yes, Joshua noted in internal chat: except ValueError:
self._distribution = scipy.stats.multivariate_normal(mean=np.zeros(self._rank), cov=np.eye(self._rank))
self._needsTransform = True because scipy/lapack are not handling a mean=0, cov=[[1]] distribution. |
@PaulTalbot-INL In the previous process, we do SVD decomposition first, and then using single variate normal distribution for each latent variable, in this case, we do not require the COV to be SPD. But in general, if you sample it directly from MVN, I think a SPD COV is required. |
This is the pull request that updated lapack in conda: conda-forge/blas-feedstock#113 |
From the other pull request larsoner said : |
So I am guessing that the new version should show up in https://anaconda.org/conda-forge/liblapack/files real soon now? |
I hope so. Even the internal |
Testing if a newer version of scipy will fix this: #2250 |
Job Mingw Test on 108be4a : invalidated by @PaulTalbot-INL lib update, rerun tests |
Checklists pass. |
Pull Request Description
What issue does this change request address? (Use "#" before the issue to link it, i.e., #42.)
#1114
What are the significant changes in functionality due to this change request?
For Change Control Board: Change Request Review
The following review must be completed by an authorized member of the Change Control Board.
<internalParallel>
to True.raven/tests/framework/user_guide
andraven/docs/workshop
) have been changed, the associated documentation must be reviewed and assured the text matches the example.