python2 iterator bug?

[tombosc]

here is the snippet:

from __future__ import print_function
from torchtext.datasets import WikiText2

i_train, i_dev, i_test = WikiText2.iters(device='-1', root="data/")
for i in i_train:
    print(i)

here's the error:

Traceback (most recent call last):
  File "bug.py", line 5, in <module>
    for i in i_train:
  File "/u/bosctom/.local/lib/python2.7/site-packages/torchtext/data/iterator.py", line 218, in __iter__
    self.batch_size - len(text)))
TypeError: can't multiply sequence by non-int of type 'float'

The bug is related to python2's math.ceil() returning a float?

[nelson-liu]

what version of torchtext are you using? A git hash would be nice, if you have one (or did you install from pip)? I get a different error related to the WikiText2 dataset not being updated to use the latest Vocab API on Python 2 and 3.

[tombosc]

I got that error using the pip version: 0.1.1

[nelson-liu]

yes, you're right --- there needs to be an int call at

text/torchtext/data/iterator.py

Line 220 in 0634892

text = text + ([TEXT.pad_token] * (math.ceil(len(text) / self.batch_size) *

. Would you be willing to create a PR, or should i go ahead?