def _classify(self, strings):
strings = [language_detection_textcleaning(i) for i in strings]
subs = [
' '.join(s)
for s in self._bpe.encode(strings, output_type = self._type)
]
transformed = self._vectorizer.transform(subs)
batch_x = _convert_sparse_matrix_to_sparse_tensor(transformed)
probs = self._sess.run(
self._softmax,
feed_dict = {self._model.X: batch_x[0], self._model.W: batch_x[1]},
)
return probs
def _classify(self, strings):
strings = [language_detection_textcleaning(i) for i in strings]
subs = [
' '.join(s)
for s in self._bpe.bpe.encode(strings, output_type=self._bpe.mode)
]
transformed = self._vectorizer.transform(subs)
batch_x = _convert_sparse_matrix_to_sparse_tensor(transformed)
r = self._execute(
inputs=batch_x,
input_labels=[
'X_Placeholder/shape',
'X_Placeholder/values',
'X_Placeholder/indices',
'W_Placeholder/shape',
'W_Placeholder/values',
'W_Placeholder/indices',
],
output_labels=['logits'],
)
probs = softmax(r['logits'], axis=-1)
return probs
def _predict(self, strings):
strings = [
language_detection_textcleaning(string) for string in strings
]
return self._model.predict(strings)