Closed kostastsa closed 1 year ago
Thanks for contributing this! I'll do a code review asap, but at first glance this looks like exactly what we need!
Thanks for getting involved!
It might be useful to separate this test from the tfp-based tests (to give e.g. test_kalman_tfp()
and test_kalman_vs_joint()
) so that we get the results from the joint test even if the tfp test fails.
If you're up for it the ideal thing might be to refactor the tests to use the class-based approach as in e.g. info_infererence_test.py and parallel_inference_test.py (described here). No worries at all if that's not possible, I'm very happy to do it after this gets merged.
+1 to splitting the test into 2 - one for TFF, one for your code.
Sounds good, I'm on it!
I have done the refactor, following the pattern used in info_inference_tests.py. Let me know how it looks!
LGTM.
I will merge despite the failing test, which is unrelated to this PR (has something to do with tqdm).
Reference #229 Implemented large MVN and added test comparing to Kalman smoother. Feedback welcome!