#11754 introduces a warning that is very confusing to me.

I get a bunch of warnings of the form:

ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 6.524430864374153e-05, tolerance: 2.30509725

What's confusing is:

• the tolerance is bigger than the duality gap, so I would assume that means convergence.
• the tolerance is not what was passed, which is 0.001
• the duality gap is smaller than the tolerance that was passed.

I have no idea what's going on here.

cc @rth @agramfort @jnothman @brentfagan @GaelVaroquaux

minimal example:

from sklearn.linear_model import LassoCV
from sklearn.datasets import load_boston
boston = load_boston()
lasso = LassoCV().fit(boston.data, boston.target)