PR: Single-stage optimization add-ons #301
Conversation
Update branch
| @@ -253,6 +258,7 @@ def _jac(self, x: RealArray = None): | |||
| x = xs[:, j] | |||
| mpi.comm_groups.bcast(x, root=0) | |||
| opt.x = x | |||
| self.opt.non_dofs = non_dofs | |||
There was a problem hiding this comment.
How is the non_dofs attribute used in the optimizable object? Is there any convention or standardization you have in mind here?
There was a problem hiding this comment.
I was thinking that it should be an array of doubles. I have no other use cases in mind.
There was a problem hiding this comment.
There was a problem hiding this comment.
This smells funny to me, because non_dofs is an attribute only of objects created by make_optimizable(). Most optimizable objects don't have a non_dofs attribute.
Update branch with latest simsopt source code
|
For a single-stage example with dofs and non-dofs, check |
|
Ready to be reviewed again |
src/simsopt/_core/optimizable.py
Outdated
| @@ -93,6 +93,8 @@ def __init__(self, | |||
| """ | |||
| if x is None: | |||
| x = np.array([]) | |||
| elif np.isscalar(x): | |||
There was a problem hiding this comment.
If this is not added the tests don't pass
There was a problem hiding this comment.
Which test exactly doesn't pass? What exactly is the error?
There was a problem hiding this comment.
This is now not needed as dofs are now handled correctly in J
|
I am not really sure why the tests are not passing, @landreman @mbkumar could you take a look? |
Codecov ReportPatch coverage:
Additional details and impacted files@@ Coverage Diff @@
## master #301 +/- ##
=======================================
Coverage 91.34% 91.35%
=======================================
Files 68 68
Lines 12009 12042 +33
=======================================
+ Hits 10970 11001 +31
- Misses 1039 1041 +2
Flags with carried forward coverage won't be shown. Click here to find out more.
☔ View full report in Codecov by Sentry. |
options for SquaredFlux definition
|
All comments addressed, now waiting for the tests to finish. |
|
Updated branch with latest SIMSOPT master branch |
rogeriojorge
force-pushed
the
rj/single_stage_PR
branch
2 times, most recently
from
c0962fe to
be922d7
Compare
|
I found a way of keeping the old dofs and non-dofs structure. The single stage example has a few changes but now the whole simsopt has fewer changes and keeps the same old dofs and non-dofs structure. Hopefully all tests pass now. |
Yet another single-stage refactor
The pull request introduces a few features to allow single-stage optimizations to be performed using SIMSOPT. In particular, these are the minimal set of changes needed to fully reproduce the results in https://arxiv.org/abs/2302.10622.
Brief list of changes: