Open tiagoleonmelo opened 3 months ago
Hey @tiagoleonmelo, Thank you for this PR, note that it seems like the unit tests are not all passed. Make sure to check your linting ;) We will take a look at the PR as soon as all the tests have passed.
Hey @tiagoleonmelo, Thank you for this PR! For the moment it doesn't pass some tests, but we have fixed the issue. So you can git merge upstream/main and everything should work! Thank you!
Hey @LacombeLouis, sorry about the delay. I synced my fork, but still running into a few issues which I don't think come from my changes - getting a 404 somewhere
Description
Allow users to pass
random_state
inMapieQuantileRegressor
constructor.When calling
fit()
, the following logic is applied:random_state
was set for this instance, and no random_state is passed in the function call, use the random state set in the constructorrandom_state
is provided neither in fit nor constructor, norandom_state
is usedrandom_state
is provided in both fit, and constructor, therandom_state
from fit is usedMy reasoning is that the
random_state
provided infit
will overwrite the one previously provided. This doesn't feel super intuitive to me (maybe the random state should not be passable now duringfit
to not have random state mismatches?), but I think it is better compared to the alternativeFixes #405
Type of change
How Has This Been Tested?
fit()
, mimicking the desired behaviorChecklist
make lint
make type-check
make tests
make coverage
make doc