Transfers the batch to CUDA.
This function is used when you don’t know what is the structure of a batch.
A tensor or an iterable of tensors.
Transferred to CUDA batch.