Is there a way for mlr3torch to make a predefined task for LSTMs? Especially for univariate time series forecasting. I saw something for LSTMs in the original torch book but I was hoping for mlr3torch to be slightly faster and to use less code. So for this I'm unsure.