def main():
max_len = 40
dataset = DataManager(path = data_dir)
dictionary = dataset.clean_train_dict
voc_size = dictionary.voc_size
dataset.BuildTrainableData(max_len = max_len)
tf.reset_default_graph()
model2 = S2VT_attention(voc_size = voc_size, max_len = max_len, dtype = tf.float32)
model2.compile()
model2.restore()
test_id = dataset.raw_data['test_id']
with open(output_dir, 'w') as f:
for i, vid in enumerate(test_id):
predict = model2.predict(dataset.test_x[i])
predict = dataset.clean_train_dict.indexlist2wordlist(predict)
sentence = ''
for word in predict:
if word == '<PAD>':
sentence = sentence.strip()
break
sentence += word + ' '
output_line = vid + ',' + sentence
if (i + 1) != len(test_id):
output_line += '\n'
f.write(output_line)
def main():
start_time = time.time()
dataset = DataManager()
# load dictionary
dict_file = open('data/dictionary.txt', 'rb')
word_list = pickle.load(dict_file)
voc_size = len(word_list)
dataset.train_dict.train_dict = word_list
dataset.train_dict.voc_size = voc_size
dictionary = dataset.train_dict
# load data from input file
dataset.LoadTestData(sys.argv[1])
test_data = np.asarray(dataset.data)
batch_size = 100
tf.reset_default_graph()
model = Seq2Seq(voc_size, batch_size=batch_size, mode='test')
model.compile()
output_file = open(sys.argv[2], 'w', encoding='utf8')
count = 0
for i in range(int(len(test_data) / batch_size)):
predict_labels = model.predict(test_data[i * batch_size:((i + 1) *
batch_size)])
for j in range(batch_size):
count += 1
result = dictionary.index2sentence(predict_labels[j][0])
if result.replace(' ', '') == '':
result = '...'
output_file.write("%s\n" % result)
while count <= len(test_data):
count += 1
output_file.write("...\n")
print('Cost time: %.2f minutes' % ((time.time() - start_time) / 60.0))
def set_testing_data(self):
print('setting testing data...')
self.max_len = 25
dataset = DataManager(max_len = self.max_len)
self.val_size, self.dictionary = dataset.getTestData()
tf.reset_default_graph()
def set_training_data(self):
print('getting training data...')
self.max_len = 25
dataset = DataManager(max_len = self.max_len)
self.val_size,self.train_x,self.train_y,self.dictionary = dataset.getTrainData()
tf.reset_default_graph()
if __name__ == '__main__':
nparts = int(sys.argv[1])
training = int(sys.argv[2])
# Initialize logger
lg = Log()
# Define data location and timesteps
trainingDir = ['../training-data/converge-diverge','../training-data/periodic-hills', \
'../training-data/square-cylinder', '../training-data/square-duct', \
'../training-data/tandem-cylinders']
trainingDir = [os.path.join(os.getcwd(), dir0) for dir0 in trainingDir]
ransTimes = [60, 90, 60, 60, 60]
lesTimes = [200, 1000, 250, 1700, 170]
dataManager = DataManager(trainingDir, ransTimes, lesTimes)
foamNN = FoamSVGD(nparts) # Number of SVGD particles
# Load pre-trained neural networks
#foamNN.loadNeuralNet('./torchNets/foamNet')
# First set up validation dataset
#foamNN.getTrainingPoints(dataManager, n_data=500, n_mb=256)
#foamNN.getTestingPoints(dataManager, n_data=500, n_mb=256)
XTdirs = ['../../IgnDelay/xdataTr']
YTdirs = ['../../IgnDelay/ydataTr']
Xdirs = ['../../IgnDelay/xdataTe']
Ydirs = ['../../IgnDelay/ydataTe']
# coding: utf-8
from utils.dataManager import DataManager
from utils.Seq2Seq import Seq2Seq
import tensorflow as tf
import numpy as np
import os
import pickle
import sys
dataset = DataManager()
# load training data and label
dataset.LoadData(file_name=sys.argv[1])
train_data = np.asarray(dataset.data)
train_label = np.asarray(dataset.label)
# load dictionary
dict_file = open('data/dictionary.txt', 'rb')
word_list = pickle.load(dict_file)
voc_size = len(word_list)
dataset.train_dict.train_dict = word_list
dataset.train_dict.voc_size = voc_size
dictionary = dataset.train_dict
# define epoch num
epoch_num = 30
tf.reset_default_graph()
model = Seq2Seq(voc_size)
model.compile()
model.fit(train_data, train_label, epoch_num)
from utils.dataManager import DataManager
from models.hw2_1_S2VT_attention import Seq2Seq
import tensorflow as tf
import numpy as np
import sys
data_dir = sys.argv[1]
max_len = 40
dataset = DataManager(path=data_dir)
dictionary = dataset.clean_train_dict
voc_size = dictionary.voc_size
dataset.BuildTrainableData(max_len=max_len)
tf.reset_default_graph()
model = Seq2Seq(voc_size=voc_size, max_len=max_len, dtype=tf.float32)
model.compile()
try:
model.restore()
except:
pass
train_x = []
train_y = []
for i, labels in enumerate(dataset.train_y):
for j, label in enumerate(labels[:5]):
train_x.append(dataset.train_x[i])
train_y.append(label)
epoch = 0