Closed ablaom closed 3 years ago
This PR does this:
fixes _reformat so that MLJ user provides matrix input in form n x p (currently user must provide a features-as-columnns table or p x n matrix) (#5)
_reformat
in case MLJ user does not provide an adjoint, they are informed that performance is not optimal and a remedy is suggested
restores some invalid MLJ interface tests (as here): The signature of MLJ model predict method (as opposed to machine predict) is predict(model, fitresult, X) not predict(fitresult, X); see https://alan-turing-institute.github.io/MLJ.jl/dev/adding_models_for_general_use/#The-predict-method-1.
predict
predict(model, fitresult, X)
predict(fitresult, X)
thanks!
This PR does this:
fixes
_reformat
so that MLJ user provides matrix input in form n x p (currently user must provide a features-as-columnns table or p x n matrix) (#5)in case MLJ user does not provide an adjoint, they are informed that performance is not optimal and a remedy is suggested
restores some invalid MLJ interface tests (as here): The signature of MLJ model
predict
method (as opposed to machinepredict
) ispredict(model, fitresult, X)
notpredict(fitresult, X)
; see https://alan-turing-institute.github.io/MLJ.jl/dev/adding_models_for_general_use/#The-predict-method-1.