def load_saved_model():
if model_type == 'seq2seq':
model = Seq2SeqIntentModel()
else:
model = MultiTaskIntentModel()
model.load(args.model_path)
return model
def load_model(self):
if self.model_type == 'seq2seq':
model = Seq2SeqIntentModel()
else:
model = MultiTaskIntentModel()
model.load(self.pretrained_model)
self.model = model
def load_model(self):
with open(IntentExtractionApi.pretrained_model_info, "rb") as fp:
model_info = pickle.load(fp)
self.model_type = model_info["type"]
self.word_vocab = model_info["word_vocab"]
self.tags_vocab = {v: k for k, v in model_info["tags_vocab"].items()}
if self.model_type == "mtl":
self.char_vocab = model_info["char_vocab"]
self.intent_vocab = {
v: k
for k, v in model_info["intent_vocab"].items()
}
model = MultiTaskIntentModel()
else:
model = Seq2SeqIntentModel()
model.load(self.pretrained_model)
self.model = model
def load_model(self):
with open(IntentExtractionApi.pretrained_model_info, 'rb') as fp:
model_info = pickle.load(fp)
self.model_type = model_info['type']
self.word_vocab = model_info['word_vocab']
self.tags_vocab = {v: k for k, v in model_info['tags_vocab'].items()}
if self.model_type == 'mtl':
self.char_vocab = model_info['char_vocab']
self.intent_vocab = {
v: k
for k, v in model_info['intent_vocab'].items()
}
model = MultiTaskIntentModel()
else:
model = Seq2SeqIntentModel()
model.load(self.pretrained_model)
self.model = model
# to_categorical one-hot格式化数据, 其中one_hot函数是转化nparray格式数据
test_y = to_categorical(test_y, dataset.label_vocab_size)
train_y = to_categorical(train_y, dataset.label_vocab_size)
train_i = one_hot(train_i, len(dataset.intents_vocab))
test_i = one_hot(test_i, len(dataset.intents_vocab))
train_inputs = [train_x, train_char]
train_outs = [train_i, train_y]
test_inputs = [test_x, test_char]
test_outs = [test_i, test_y]
######################################################################
print('Building model')
model = MultiTaskIntentModel(use_cudnn=args.use_cudnn)
model.build(dataset.word_len,
dataset.label_vocab_size,
dataset.intent_size,
dataset.word_vocab_size,
dataset.char_vocab_size,
word_emb_dims=args.token_emb_size,
tagger_lstm_dims=args.lstm_hidden_size,
dropout=args.tagger_dropout)
# initialize word embedding if external model selected
# 如果存在词向量model，就初始化词向量
if args.embedding_model is not None:
print('Loading external word embedding')
embedding_model, _ = load_word_embeddings(args.embedding_model)
embedding_mat = get_embedding_matrix(embedding_model,