fix transform inits input and output sizes

[SteveBronder]

Fix for transform inits so that the input vector is the same as the number of constrained parameters while the output vector is the size of the number of unconstrained params

Submission Checklist

• Run unit tests
• Documentation
  • If a user-facing facing change was made, the documentation PR is here:
  • OR, no user-facing changes were made

Release notes

Copyright and Licensing

By submitting this pull request, the copyright holder is agreeing to
license the submitted work under the BSD 3-clause license (https://opensource.org/licenses/BSD-3-Clause)

[WardBrian]

OCaml looks fine to me, can't comment much on the generated C++/whether this resolves the issue. I assume you tried it out?

[SteveBronder]

Yeah this fixes the issue in #992, tested out that model. As the summary says that goof happened because of the difference in the number of constrained and unconstrained parameters for things like simplex and unit vectors

[wds15]

As this PR is scheduled to make up a 2.28.1... I would ask to hold for a moment that release so that we also fix the issues seen in the Adjoint ODE solver if possible... at least I would like to try a debug session today and if it turns out that the problem can be fixed in finite time, then that would ideally also be part of 2.28.1?

[rok-cesnovar]

Yeah, definitely, we can wait. Steve will bring up the id=1 thing (event though I think we have come to an agreement) on today's meeting so we were planning on holding off at least until after the meeting. Will follow the Discourse thread for updates on the adjoint ODE issue. If you think the issue is too big for quick fix, we will release 2.28.1 with this PR.

[wds15]

Yup... I don't want to delay too much, but if there is a chance to include a fix, then that's great. If not we go for 2.28.2 or we see.

[SteveBronder]

@rok-cesnovar if you approve I think this is good!