[documentation] Explain coloring and detector kwargs in sparse backends #249
Conversation
|
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #249 +/- ##
==========================================
- Coverage 95.48% 95.44% -0.04%
==========================================
Files 13 13
Lines 1439 1427 -12
==========================================
- Hits 1374 1362 -12
Misses 65 65 ☔ View full report in Codecov by Sentry. |
There was a problem hiding this comment.
Looks good to me, but in general it might be useful to add more (@ref) and (@extref) (from DocumenterInterLinks.jl)
docs/src/mixed.md
Outdated
| @@ -1,4 +1,4 @@ | |||
| # Build a hybrid NLPModel | |||
| # Build an hybrid NLPModel | |||
There was a problem hiding this comment.
nope, it was correct before
| H = hess(nlp, x) | ||
| ``` | ||
|
|
||
| The available backends for sparse derivatives (`SparseADJacobian`, `SparseADHessian` and `SparseReverseADHessian`) have keyword arguments `detector` and `coloring` to specify the sparsity pattern detector and the coloring algorithm, respectively. |
There was a problem hiding this comment.
It's rather weird to call it SparseADHessian versus SparseReverseADHessian, a better name would be forward-over-forward vs. forward-over-reverse if I understand correctly. But with a switch to DI those distinctions will disappear
There was a problem hiding this comment.
I didn't know that Tangi implement SparseReverseADHessian.
The name is not good, we should switch to DI.jl asap!
amontoison
merged commit c1b0919
into
JuliaSmoothOptimizers:main
Related to #247
@gdalle
I renamed the keyword argument for the coloring algorithm, but I believe it's safe to do so.
Since this option was not documented previously, I don't expect any users to rely on the
algoption.