Meaning of "iterations" in readme about pointer_summarizer HOT 2 CLOSED

I think iteration is for the number of batch updates. You can find it from Train.trainIters() from train.py.

To relate this iteration to the concept of epoch, you need to see how they make the batches, which you can find from example_generator() in data.py. From the data_path, it keeps pulling the batches from data_path until they consume all files in the path. After that, they redo this generation process over the data_path again (since it is a generator) unless argument single_pass is set as True.

In summary, (total # of examples/batch_size) iterations would be approximately the same as one epoch, but note that for the last batch it may have mixed examples.

from pointer_summarizer.

Thanks dear! I see. I close the issue

from pointer_summarizer.