GRU on Time-Series Data

I am running a GRU on a time-series dataset to predict the outcomes. I have cut the dataset into sequence lengths of 200 and have run a batch size of 128. I was wondering, after training the dataset using DataLoader, is there an easy way to piece the different outcomes?

I am trying to compare the predictions and the actual data on the whole dataset but am unsure how to piece everything together (since the batches are divided randomly, and are not tagged with the correct date-time index).

My question is: is there a way to ‘piece back the data’ after using the dataloader to divide the data into randomized iterable batches?