2018-01-23 18:11:53 +00:00
|
|
|
'''Test that models with no pretrained vectors can be deserialized correctly
|
|
|
|
|
after vectors are added.'''
|
2018-01-23 18:11:01 +00:00
|
|
|
from __future__ import unicode_literals
|
|
|
|
|
import numpy
|
|
|
|
|
from ...pipeline import Tagger
|
|
|
|
|
from ...vectors import Vectors
|
|
|
|
|
from ...vocab import Vocab
|
|
|
|
|
from ..util import make_tempdir
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def test_issue1727():
|
|
|
|
|
data = numpy.ones((3, 300), dtype='f')
|
|
|
|
|
keys = [u'I', u'am', u'Matt']
|
|
|
|
|
vectors = Vectors(data=data, keys=keys)
|
|
|
|
|
tagger = Tagger(Vocab())
|
|
|
|
|
tagger.add_label('PRP')
|
|
|
|
|
tagger.begin_training()
|
|
|
|
|
|
|
|
|
|
assert tagger.cfg.get('pretrained_dims', 0) == 0
|
|
|
|
|
tagger.vocab.vectors = vectors
|
|
|
|
|
|
|
|
|
|
with make_tempdir() as path:
|
|
|
|
|
tagger.to_disk(path)
|
|
|
|
|
tagger = Tagger(Vocab()).from_disk(path)
|
|
|
|
|
assert tagger.cfg.get('pretrained_dims', 0) == 0
|
