-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
create ApproximateGPs.TestUtils #117
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for making some headway on this -- it will be good to have this formalised a bit more. Hopefully it will let us move towards a better understanding of scope / limitations of this particular package.
One thing I notice is that the approx_lml
function isn't tested, which you point to in the other PR. Should this at least be checked?
Re. my comment about separating out the API from the ability to provide good approximate inference, should we also consider including applying Zygote to approx_lml
as part of this test suite somewhere? (I'm not committed to requiring this though).
Yeah, we should have that too... maybe later? for now I just wanted to take out the non-EP changes from #64.
Testing gradients should probably be there, too ,but also separately (e.g. current EP doesn't have approx_lml implemented, and gettings its gradients right would be another headache on top) |
Happy to leave these suggestions for another PR. Please ping me again when you want me to take another look -- I think it's pretty close to being good to go. |
…eGPs.jl into st/testutils
@willtebbutt ping |
Codecov Report
@@ Coverage Diff @@
## master #117 +/- ##
==========================================
- Coverage 94.13% 94.13% -0.01%
==========================================
Files 4 5 +1
Lines 290 324 +34
==========================================
+ Hits 273 305 +32
- Misses 17 19 +2
Continue to review full report at Codecov.
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM subject to formatting suggestion
didn't think; this won't work This reverts commit 37a541e.
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
To make it easier to test various approximations (similar to AbstractGPs.TestUtils)