Create a dataset from sequential data PyTorch

I have a dataset with IMU information at each timestep and the corresponding ground truth location. I want to create a dataset to be fed into a RNN network in PyTorch but I do not know how to create the Dataset class as the time order needs to be kept and a batch of items is considered. So consider I have many CSV files with 4 columns (i.e. orientation x, orientation y, acceleration x, acceleration y) and 10k timesteps each and other ground truth files with 10k timesteps and the real location for each timestep in each file.

I have already accomplish to get IMU batches of 200 samples from all the files, so the dimensions are [total number of subdivisions or batches, 200, 4]. Now I do not know how to create a dataset to be used by PyTorch to iterate over multiple batches at the same time (i.e. equal to when doing Sequence-to-Sequence in Text where each batch contains several sentences).