Once you have calculated Mean Absolute Errors and Mean Squared Errors in each fold of all the models, you would choose the model having the minimum errors as well as the models with minimum variance in errors i.e. the model with minimum mean and std of Mean Absolute Errors and Mean Squared Errors. You will be further fine-tuning this chosen model.

Which of these models you will select for fine-tuning (Grid Search) based on the mean RMSE values?