Batch

The Batch class is an internal data representation loaded by DataLoader and input into the prediction model.

The objects retrieved from the Dataloader of the data flow are all objects of the Batch class.

The Batch class is an abstract data structure based on the implementation of python.dict, which constitutes the structure of key-value, where key is the feature name, and value is the tensor(torch.Tensor) corresponding to the feature name, which contains all the data of a batch or mini-batch.

So you can use the following methods to get the corresponding value:

loc = batch['current_loc']
tim = batch['current_tim']

The Batch class unifies the input format of the model so that the framework can implement general executor classes and standard model classes.