[Python] why TensorDataset devide the data to minibatches?

Why TensorDataset devide the data to minibatches? For example, when putting in it 2D array, instead of yielding 2D tensors as batches, it sets the required batches to be minibatches, and its actual "batch" size is a list of length 1. For example, why in this example the batch size is 1 and not 3?

from torch.utils.data import TensorDataset

data = torch.randint(0, 100, (20, 3), dtype=torch.int32)
tensor_dataset = TensorDataset(data)

print('len(tensor_dataset)', len(tensor_dataset)) # output: 20
for batch in tensor_dataset:
print('type(batch)', type(batch)) ## output: tuple
print('len(batch)', len(batch)) ## output: 1 <<<******** Why not 3? ********
for minibatch in batch:
print('type(minibatch)', type(minibatch)) ## output: tensor
print('len(minibatch)', len(minibatch)) ## output: 3
break
break


Is there built-in solution for this, or it must be handled manually by unpacking the batches to minibatches in the dataloader loop?

Continue reading...