def def_font2char2word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
word_array = qndex.nlp.def_word_array()
def model(document, label=None, *, mode, key=None):
return classify(
font2char2word2sent2doc(
document,
words=word_array(),
mode=mode,
fonts=font_array(),
**adder.flags),
label,
key=key,
predictions={
'font_attentions': tf.tile(
tf.expand_dims(
collections.get_attentions()[0],
axis=0),
[tf.shape(document)[0], 1, 1, 1]),
},
mode=mode,
regularization_scale=qnd.FLAGS.regularization_scale)
return model
def def_word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
def model(document, label=None, *, mode):
return classify(word2sent2doc(document,
word_space_size=len(qnd.FLAGS.words),
**adder.flags),
label,
mode=mode)
return model
def def_word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
get_words = qndex.nlp.def_words()
def model(document, label=None, *, mode, key=None):
return classify(word2sent2doc(document,
word_space_size=len(get_words()),
**adder.flags),
label,
key=key,
mode=mode)
return model
def def_char2word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
word_array = qndex.nlp.def_word_array()
def model(document, label=None, *, mode):
return classify(char2word2sent2doc(document,
words=word_array(),
char_space_size=len(
qnd.FLAGS.chars),
**adder.flags),
label,
mode=mode)
return model
def def_word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
get_words = qndex.nlp.def_words()
def model(document, label=None, *, mode, key=None):
return classify(
word2sent2doc(
document,
word_space_size=len(get_words()),
**adder.flags),
label,
key=key,
mode=mode,
regularization_scale=qnd.FLAGS.regularization_scale)
return model
def def_font2char2word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
word_array = qndex.nlp.def_word_array()
def model(document, label=None, *, mode, key=None):
return classify(font2char2word2sent2doc(document,
words=word_array(),
mode=mode,
fonts=font_array(),
**adder.flags),
label,
mode=mode,
key=key,
regularization_scale=qnd.FLAGS.regularization_scale)
return model
def def_char2word2sent2doc():
adder = add_flags()
adder.add_flag("char_embedding_size", type=int, default=100)
classify = qndex.def_classify()
word_array = qndex.nlp.def_word_array()
def model(document, label=None, *, mode, key=None):
return classify(
char2word2sent2doc(
document,
words=word_array(),
char_space_size=len(qnd.FLAGS.chars),
**adder.flags),
label,
key=key,
mode=mode,
regularization_scale=qnd.FLAGS.regularization_scale)
return model
def def_font2char2word2sent2doc():
adder = add_flags()
classify = qndex.def_classify()
word_array = qndex.nlp.def_word_array()
def model(document, label=None, *, mode, key=None):
return classify(font2char2word2sent2doc(document,
words=word_array(),
mode=mode,
fonts=font_array(),
**adder.flags),
label,
key=key,
predictions={
'font_attentions':
tf.tile(
tf.expand_dims(collections.get_attentions()[0],
axis=0),
[tf.shape(document)[0], 1, 1, 1]),
},
mode=mode,
regularization_scale=qnd.FLAGS.regularization_scale)
return model