February 2018
Intermediate to advanced
262 pages
6h 59m
English
Let's take a look at the code and understand the two key things involved in the batching of sequential data:
train_iter, valid_iter, test_iter = data.BPTTIterator.splits( (train, valid, test), batch_size=20, bptt_len=35, device=0)
There are two important things that are going through this method. One is batch_size, and another is bptt_len, called backpropagation through time. It gives a brief idea about how data is transformed through each phase.
Read now
Unlock full access