Skip to content

Conversation

jessegrabowski
Copy link
Member

I was playing around with DADVI and wanted the full range of optimizer options available in find_MAP, so I added them. These include:

  • compile_kwargs
  • gradient_backend (for MIT-MOT scan graphs in JAX mode until we fix that)
  • basinhopping
  • Results are now returned with a posterior group like all other PyMC inference methods
  • If include_transform = True, the unconstrained draws are returned in an unconstrained_posterior group
  • The optimizer_results are returned as an idata group

I also added tests for DADVI, mostly copy/paste from the find_MAP tests.

The result is some amount of code duplication with find_MAP, but I think it's acceptable.

cc @martiningram

Copy link

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.


Powered by ReviewNB

@jessegrabowski
Copy link
Member Author

@maresb looks like the BLAS installation is failing on the new windows machines, is that known to us?

Copy link
Member

@ricardoV94 ricardoV94 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The CI failing is unrelated to this PR, so I think it's fine to go ahead and merge

@jessegrabowski jessegrabowski merged commit 37f4588 into pymc-devs:main Sep 23, 2025
10 of 17 checks passed
@jessegrabowski jessegrabowski deleted the dadvi-options branch September 23, 2025 16:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancements New feature or request tests
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants