refactoring: in LogLinearModel::Train() the local variable gradient (as well as appliedL1Penalties) does not need to be an instance of LogLinearParams. It doesn't make much of a difference memory-wise, but it will be less confusing. #12
Open
{{ message }}
