com.intel.analytics.bigdl.dataset
total batch size
feature padding value (by default None, meaning no feature padding)
label padding value (by default None, meaning no label padding)
if padding, it specifies the length of feature/label after padding (by default None, meaning the length after padding is set to the max length of feature/label in a mini-batch)
partition number of dataset, default means partitionNum equals Engine.nodeNumber()
Convert a sequence of Sample to a sequence of MiniBatch, optionally padding all the features (or labels) in the mini-batch to the same length