Garrafao / durel_system_annotators

3 stars 0 forks source link

update model testing #22

Closed Garrafao closed 6 months ago

Garrafao commented 8 months ago

@AinaIanemahy Currently, tests/data.py is missing. On purpose?

Garrafao commented 8 months ago

tests/data.py is now there. With commit 8a230da it extracts binary test data and data with multiple labels (median), e.g.:

Important: you need to rerun data.py for the current integration tests to work with the new data format.

Garrafao commented 8 months ago

@shafqatvirk I think that there should be two integration test files, one per data format (_binarize-median, _median), see above. On both data types we can test all models using Spearman for evaluation. On _median we can additionally evaluate only the binary annotator with accuracy.

Each of the two integration test files could contain the tests for all four models, as you had proposed before.

shafqatvirk commented 8 months ago

So we should have integration tests for all three 'random', 'single_threshold', and multi_threshold models with the above two data sets (_binarize_median and _median)?

Garrafao commented 7 months ago

Integration tests should be adjusted to the new file format as described in #43.

AinaIanemahy commented 6 months ago

All integration tests have been run, so I think that we can close this issue.