changed regressions and weights tests to use API#480
Conversation
|
Looks good to me at first sight. |
| columns=map(str, | ||
| range(th.shape[1]))) | ||
| def test_predictions(data_config): | ||
| #TODO: change the baseline to just be the `experiment_result_table` |
There was a problem hiding this comment.
I didn't understand what is the problem with doing this.
There was a problem hiding this comment.
so if I were to have:
exp_res_tab = API.experiment_result_table_no_table(**data_config)
return exp_res_tab.iloc[:, 2:]then the indexes don't match. I just wanted to make sure that the tables matched before changing the baseline table... Although now that I think about it, it would have been apparent on the git diff anyway..
There was a problem hiding this comment.
The solution to that is really #449 in some form...
There was a problem hiding this comment.
Yeah sure, actually I played with it for a bit and I think this might be the best solution for now, because the tableloader which saves the table sane_load by default needs 1 level of indexing and so to mess around with making new indexes is more effort than to just take the values and construct a new dataframe as is being done here.
Maybe I should expand the comment slightly but then leave it as is?
…d reconstruct a pd.DataFrame due to limitations of CSV
|
Ok I think I'm happy with this |
| covs = [results.experiment_covariance_matrix(exp, False, None) for exp in exps] | ||
| return covs | ||
| def data_witht0_config(): | ||
| experiment_list = [ |
There was a problem hiding this comment.
Does it make sense to move the common experiment_list outside?
There was a problem hiding this comment.
Yeah actually these configs are all copy pastas which is probably unneccessary I'll trim it down
|
oops there are some unused imports now |
|
Hello, this is @Zaharid's automated QA script. Please note it is highly experimental. I ran pylint on your changes and found some new issues. On
|
2 participants
closes #459
There are a couple of tests which I wanted to show can produce the same numbers in this commit but I think instead of creating a pd.DataFrame which we take values from to create a different one, the CSVs for relevant tests could just be changed to be the original CSVs (this would just be changing columns/indices) - These are marked with a TODO and should be done before merging
Most of the tests become simpler, there is the caviat that the fits test needs a work around which kind of defeats the point of the API but I felt like deleting as much as I could from
conftest.pycriticism welcome..