class Trainset(Dataset):
PREFIX = 'train'
class Parseset(Dataset):
PREFIX = 'parse'
#***************************************************************
if __name__ == '__main__':
""" """
from nparser.vocabs import *
from nparser.dataset import Trainset
configurable = Configurable()
dep_vocab = DepVocab.from_configurable(configurable)
word_vocab = WordVocab.from_configurable(configurable)
lemma_vocab = LemmaVocab.from_configurable(configurable)
pretrained_vocab = PretrainedVocab.from_vocab(word_vocab)
char_vocab = NgramMultivocab.from_vocab(word_vocab)
word_multivocab = Multivocab.from_configurable(
configurable, [word_vocab, pretrained_vocab, char_vocab], name='words')
tag_vocab = TagVocab.from_configurable(configurable)
xtag_vocab = XTagVocab.from_configurable(configurable)
head_vocab = HeadVocab.from_configurable(configurable)
rel_vocab = RelVocab.from_configurable(configurable)
trainset = Trainset.from_configurable(configurable, [
dep_vocab, word_multivocab, lemma_vocab, tag_vocab, xtag_vocab,
head_vocab, rel_vocab
])
matrix = self.embeddings.eval()
with codecs.open(self.name+'.txt', 'w') as f:
for idx in range(self.START_IDX, len(self)):
f.write('%s %s\n' % (self[idx], ' '.join(matrix[idx])))
return
#=============================================================
@property
def pretrained_vocab(self):
return self._pretrained_vocab
#=============================================================
def __setattr__(self, name, value):
if name == '_pretrained_vocab':
self._str2idx = value._str2idx
self._idx2str = value._idx2str
self._counts = value._counts
super(RetrainedVocab, self).__setattr__(name, value)
#***************************************************************
if __name__ == '__main__':
""" """
from nparser import Configurable
from nparser.vocabs import PretrainedVocab
configurable = Configurable(retrained_vocab={'embed_loss':'cross_entropy', 'retrained_embed_size':50})
pretrained_vocab = PretrainedVocab.from_configurable(configurable)
retrained_vocab = RetrainedVocab.from_vocab(pretrained_vocab)
retrain_loss = retrained_vocab(pretrained_vocab)
print('RetrainedVocab passes',file=sys.stderr)