def main(_):
word_to_id, word_embedding = reader.load_vocabulary(word_embedding_path)
train_data_producer = reader.DataProducer(word_to_id, train_path)
valid_data_producer = reader.DataProducer(word_to_id, valid_path, False)
graph = bigru.BiGRU(len(word_embedding), word_embedding)
graph.train(train_data_producer, valid_data_producer, 100)
def main(_):
word_to_id, word_embedding = reader.load_vocabulary(word_embedding_path)
test_data_producer = reader.DataProducer(word_to_id, test_path, False)
model_path = '../save_models/bigru2layers_crf_50472'
graph = bigru2layers_crf.BiGRU2LayersCRF(len(word_embedding),
2,
word_embedding,
batch=1)
print graph.evaluate(test_data_producer, model_path)
def main(_):
id2tagsinwords_map = reader.buildID2TagsInWordsMap(relation_data_dir)
word_to_id, word_embedding = reader.load_vocabulary(word_embedding_path)
relation_embedding = reader.load_relation_embeddings(
relation_embedding_path)
train_data_producer = reader.DataProducer(id2tagsinwords_map, word_to_id,
train_path, 1024)
valid_data_producer = reader.DataProducer(id2tagsinwords_map, word_to_id,
valid_path, 1024, False)
graph = bigru2layers_dev.BiGRU2LayersDev(len(word_embedding),
len(relation_embedding),
word_embedding,
relation_embedding)
graph.train(train_data_producer, valid_data_producer, 100)
def main(_):
id2tagsinwords_map = reader.buildID2TagsInWordsMap(relation_data_dir)
word_to_id, word_embedding = reader.load_vocabulary(word_embedding_path)
relation_embedding = reader.load_relation_embeddings(
relation_embedding_path)
test_data_producer = reader.DataProducer(id2tagsinwords_map, word_to_id,
test_path, 1024, False)
# BiGRU2LayersDev
model_path = '../save_models/bigru2layers_dev_14421'
graph = bigru2layers_dev.BiGRU2LayersDev(len(word_embedding),
len(relation_embedding),
word_embedding,
relation_embedding,
batch=1)
print graph.evaluate(test_data_producer, model_path)
def __init__(self, graph_def_path, vacab_path):
prefix = "import/"
self.input_word_ids = prefix + "Test/Model/input_word_ids:0"
self.model_length = prefix + "Test/Model/seq_len:0"
self.state_in = prefix + "Test/Model/Model/state_in:0"
self.state_out = prefix + "Test/Model/Model/state_out:0"
self.top_k = prefix + "Test/Model/top_k:0"
self.model_probs = prefix + "Test/Model/Model/probabilities:0"
self.model_top_k_result = prefix + "Test/Model/Model/top_k_prediction:1"
with open(graph_def_path, "rb") as f:
graph_def = tf.GraphDef()
graph_def.ParseFromString(f.read())
tf.import_graph_def(graph_def)
self.word_to_id, self.id_to_word = reader.load_vocabulary(vacab_path)
import os
import reader
import tensorflow as tf
import numpy as np
app = Flask(__name__)
relation_data_dir = '../../data/relation'
model_path = '../save_models/bigru2layers_dev_14421'
word_embedding_path = '../../../data/glove/glove.6B.300d.txt'
relation_embedding_path = '../../data/transE/relation_embeddings.txt'
os.environ['CUDA_VISIBLE_DEVICES'] = ''
id2tagsinwords_map = reader.buildID2TagsInWordsMap(relation_data_dir)
word_to_id, word_embedding = reader.load_vocabulary(word_embedding_path)
relation_embedding = reader.load_relation_embeddings(relation_embedding_path)
graph = bigru2layers_dev.BiGRU2LayersDev(len(word_embedding),
len(relation_embedding),
word_embedding,
relation_embedding,
batch=1)
#graph = bigru.BiGRU(len(word_embedding), len(relation_embedding),
# word_embedding, relation_embedding, batch=1)
#graph = bigru2layers.BiGRU2Layers(len(word_embedding), len(relation_embedding),
# word_embedding, relation_embedding, batch=1)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver()
if not is_special:
common_words |= set(alphanumerics)
# with open(COMMON_WORDS_SAVEPATH, 'w') as common_file:
# for w in common_words:
# common_file.write(w + '\n')
#
# with open(SPECIAL_WORDS_SAVEPATH, 'w') as special_file:
# for w in special_words:
# special_file.write(w + '\n')
return special_words, common_words
if __name__ == '__main__':
vocabulary = load_vocabulary()
unique_special_words = set()
unique_common_words = set()
#########################
##### Training Data #####
#########################
## load tokens extracted from train text
with open(config.train_tokens_path) as text_file:
text_dict = json.load(text_file)
print('complete loading processed training text')
special_words, common_words = extract_special_words_from_docs(text_dict, vocabulary)
## save common and special words extracted
with open(config.special_words_train_savepath, 'w') as output_file:
json.dump(special_words, output_file, indent=2)
