Inference GARCH(1,1) vs Forecast

Hi

I have made a neural network to predict log_returns and I assess the performance using RMSE and MAE. I want to use a GARCH(1,1) as a benchmark to compare the performance of the RMSE and MAE estimates of the neural network. I have estimated a GARCH on my in-sample data and inferred the conditional variance on the out-of-sample data. I then calculate the RMSE and MAE. My question is, is it possible to compare the RMSE and MAE of the forecasts from the neural network with those of the GARCH inference? Or do I need to forecast using the GARCH model and then calculate the RMSE and MAE?

Thanks