In scope of this issue-branch, the data handler will be updated.
In addition to a clean-up of the related methods, the data handler will be extended to handle invariant data which does not have a time-dimension in the data files. During the batch-generation however, the time-dimension must be added as long as the trained model has a single input-stream. In case of two inpit streams, the data handler must also provide two data streams.
Both cases will be tackled, whereas the latter will allow that the two input data streams are defined on different underlying grids, i.e. a coarse- and a fine-grained grid for dynamics and static predictor variables.
In scope of this issue-branch, the data handler will be updated. In addition to a clean-up of the related methods, the data handler will be extended to handle invariant data which does not have a time-dimension in the data files. During the batch-generation however, the time-dimension must be added as long as the trained model has a single input-stream. In case of two inpit streams, the data handler must also provide two data streams.
Both cases will be tackled, whereas the latter will allow that the two input data streams are defined on different underlying grids, i.e. a coarse- and a fine-grained grid for dynamics and static predictor variables.