One step forecast model testing and validation

I have a question of how to test 1-step model (for ex MLP/LSTM). I have 1 year of learning data to fit and 10 day history window and 1 step-one day forecast.

How can I test it on a new test data? Can I apply walk through validation and RMSE, MAPE and other metrics to the final "prediction curve" which is a summ of independent predictions.

Because that are different forecasts and they are not dependent from each other like in the case of recursive multistep or direct forecast.

Thanx.