Support Mixing AD Frameworks for LogDensityProblems and the objective
#180
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
Benchmark Results
| Benchmark suite | Current: 3bb03f9 | Previous: d160060 | Ratio |
|---|---|---|---|
normal/RepGradELBO + STL/meanfield/Zygote |
3657107374 ns |
3911593378.5 ns |
0.93 |
normal/RepGradELBO + STL/meanfield/ReverseDiff |
1102349821 ns |
1126577916 ns |
0.98 |
normal/RepGradELBO + STL/meanfield/Mooncake |
1134389980 ns |
1188520969 ns |
0.95 |
normal/RepGradELBO + STL/fullrank/Zygote |
3638817421.5 ns |
3851083566.5 ns |
0.94 |
normal/RepGradELBO + STL/fullrank/ReverseDiff |
1579876196 ns |
1611527605.5 ns |
0.98 |
normal/RepGradELBO + STL/fullrank/Mooncake |
1214828309 ns |
1248235339 ns |
0.97 |
normal/RepGradELBO/meanfield/Zygote |
2547816176 ns |
2738634121.5 ns |
0.93 |
normal/RepGradELBO/meanfield/ReverseDiff |
756326907 ns |
770435862 ns |
0.98 |
normal/RepGradELBO/meanfield/Mooncake |
1021460616 ns |
1062623223 ns |
0.96 |
normal/RepGradELBO/fullrank/Zygote |
2585227666.5 ns |
2780952483.5 ns |
0.93 |
normal/RepGradELBO/fullrank/ReverseDiff |
933675264.5 ns |
958044713 ns |
0.97 |
normal/RepGradELBO/fullrank/Mooncake |
1099030538 ns |
1142753555 ns |
0.96 |
normal + bijector/RepGradELBO + STL/meanfield/Zygote |
5284409807 ns |
5608097016 ns |
0.94 |
normal + bijector/RepGradELBO + STL/meanfield/ReverseDiff |
2342704841 ns |
2448081612 ns |
0.96 |
normal + bijector/RepGradELBO + STL/meanfield/Mooncake |
3869237236 ns |
3958456262.5 ns |
0.98 |
normal + bijector/RepGradELBO + STL/fullrank/Zygote |
5440495726 ns |
5573088994 ns |
0.98 |
normal + bijector/RepGradELBO + STL/fullrank/ReverseDiff |
3027567497 ns |
3058706489 ns |
0.99 |
normal + bijector/RepGradELBO + STL/fullrank/Mooncake |
4033537126 ns |
4098676416.5 ns |
0.98 |
normal + bijector/RepGradELBO/meanfield/Zygote |
4188547956.5 ns |
4289915836.5 ns |
0.98 |
normal + bijector/RepGradELBO/meanfield/ReverseDiff |
2011523681 ns |
2073166583 ns |
0.97 |
normal + bijector/RepGradELBO/meanfield/Mooncake |
3719413465.5 ns |
3800475334 ns |
0.98 |
normal + bijector/RepGradELBO/fullrank/Zygote |
4302378001.5 ns |
4419557304 ns |
0.97 |
normal + bijector/RepGradELBO/fullrank/ReverseDiff |
2287039896 ns |
2342946249 ns |
0.98 |
normal + bijector/RepGradELBO/fullrank/Mooncake |
3899698655 ns |
3965463115.5 ns |
0.98 |
This comment was automatically generated by workflow using github-action-benchmark.
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
|
This PR is superseded by a cleaned-up version #187 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR enables mixing AD frameworks for the target
LogDensityProblemand differentiating through the variational objective. This is enabled by defining a customrrulethroughChainRulesCore. As such, we are restricted to AD frameworks that support importingrrules. However, we still support directly differentiating throughLogDensityProblems.logdensityif the targetLogDensityProblemonly has zeroth-order capability.