def __init__(self, dataset_folder, dataset_file, dict_file, stopwords_file, word_embedding_file,
train_rate, valid_rate, test_rate, algo_name, charset, mode) :
"""
Need to set these attributes.
1. conf_dict: configuration of the model.
2. cr: CorpursReader for operate data.
3. model: the network model.
"""
self.conf_dict = {'algo_name': algo_name, 'batch_size': 96,
'train_valid_test_rate': [train_rate, valid_rate, test_rate],
'split_level': 'zi', 'pre_word_embedding': False,
'word_embedding_dim': 128, 'max_sentence_word_num': 150,
'min_sentence_word_num': 1, 'is_BEG': False, 'is_END': True,
'hidden_dim': 1024, 'n_style': 4,
'style_embedding_dim': 128, 'charset': charset, 'shuffle': False,
'save_freq': 100}
if mode == 'train' :
self.param_path = \
os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.model')
else :
self.param_path = \
os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.model.final')
param_dict = load_params_val(self.param_path)
self.conf_path = os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.conf')
save_confs_val(self.conf_dict, self.conf_path)
# set corpus reader
if mode == 'train' :
self.cr = CorpusReaderDialog(dataset_file=dataset_file,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
else :
self.cr = CorpusReaderDialog(dataset_file=None,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
# set model
self.model = StyleEncoderDecoderNetwork(n_words=len(self.cr.get_word_dictionary()),
hidden_status_dim=self.conf_dict['hidden_dim'],
word_embedding_dim=self.conf_dict['word_embedding_dim'],
n_style=self.conf_dict['n_style'],
style_embedding_dim=self.conf_dict['style_embedding_dim'],
input_params=param_dict)
def __init__(self, dataset_folder, dataset_file, dict_file, stopwords_file, word_embedding_file,
train_rate, valid_rate, test_rate, algo_name, charset, mode) :
"""
Need to set these attributes.
1. conf_dict: configuration of the model.
2. cr: CorpursReader for operate data.
3. model: the network model.
"""
self.conf_dict = {'algo_name': algo_name, 'batch_size': 256,
'train_valid_test_rate': [train_rate, valid_rate, test_rate],
'split_level': 'zi', 'pre_word_embedding': False,
'word_embedding_dim': 128, 'max_sentence_word_num': 150,
'min_sentence_word_num': 1, 'is_BEG': False, 'is_END': True,
'hidden_dim': 512, 'charset': charset, 'shuffle': False,
'save_freq': 100}
if mode == 'train' :
self.param_path = \
os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.model')
else :
self.param_path = \
os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.model.final')
self.param_path = 'ChoEncoderDecoder_bda37ef460ea58d4cfaf1122e4a7e2d8.model302'
param_dict = load_params_val(self.param_path)
self.conf_path = os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.conf')
#save_confs_val(self.conf_dict, self.conf_path)
# set corpus reader
if mode == 'train' :
self.cr = CorpusReaderDialog(dataset_file=dataset_file,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
else :
self.cr = CorpusReaderDialog(dataset_file=None,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
# set model
self.model = RnnEncoderDecoderNetwork(n_words=len(self.cr.get_word_dictionary()),
hidden_status_dim=self.conf_dict['hidden_dim'],
word_embedding_dim=self.conf_dict['word_embedding_dim'],
input_params=param_dict)
def train(self):
"""
Train a model.
"""
n_trainset, n_validset, n_testset = self.cr.get_size()
n_batches = max(1, (n_trainset - 1) / self.conf_dict['batch_size'] + 1)
# print model information
print '#' * 100
self.print_model_info()
print ('Compiling model...')
train_model = \
self.model.get_training_function(self.cr, batch_size=self.conf_dict['batch_size'])
valid_model = self.model.get_validing_function(self.cr, batch_size=self.conf_dict['batch_size'])
test_model = self.model.get_testing_function(self.cr, batch_size=self.conf_dict['batch_size'])
# start train
print '#' * 100
print ('Start to train.')
test_error = test_model()[0]
print ('Now testing model. Cost Error: %.10f' % (test_error))
epoch, it, n_epochs = 0, 0, 1000000
while (epoch < n_epochs):
epoch += 1
for i in xrange(n_batches):
# train model
train_error = train_model(i)[0]
# print 'Step error: %f\r' % train_error,
if math.isnan(train_error):
print ('Train error is NaN in iteration %d, batch %d' % (epoch, i))
error_model_path = self.param_path + str(epoch) + '.error'
save_params_val(error_model_path, self.model.get_parameters())
print ('model saved in %s , reload and skip the %d batch.' % \
(error_model_path, self.conf_dict['save_freq']))
param_dict = load_params_val(self.param_path)
self.model.set_parameters(param_dict)
exit()
if it % self.conf_dict['save_freq'] == 0:
# valid model
# valid_error = valid_model()[0]
print ('@iter: %d\tTraining Error: %.10f' % (it, train_error))
print self.param_path
save_params_val(self.param_path, self.model.get_parameters())
it = it + 1
# test model
# print ('Finished a epoch.')
test_error = test_model()[0]
print ('Now testing model. Testing Error: %.10f' % (test_error))
save_params_val(self.param_path + str(epoch), self.model.get_parameters())
def __init__(self, dataset_folder, dataset_file, dict_file, stopwords_file, word_embedding_file,
train_rate, valid_rate, test_rate, algo_name, charset, mode) :
"""
Need to set these attributes.
1. conf_dict: configuration of the model.
2. cr: CorpursReader for operate data.
3. model: the network model.
"""
self.conf_dict = {'algo_name': algo_name, 'batch_size': 128,
'train_valid_test_rate': [train_rate, valid_rate, test_rate],
'split_level': 'zi', 'pre_word_embedding': False,
'word_embedding_dim': 128, 'max_sentence_word_num': 150,
'min_sentence_word_num': 1, 'is_BEG': False, 'is_END': True,
'hidden_dim': 1024, 'charset': charset, 'shuffle': False,
'save_freq': 100}
if mode == 'train' :
self.param_path = \
os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.model')
else :
self.param_path = \
os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.model.final')
param_dict = load_params_val(self.param_path)
self.conf_path = os.path.join(dataset_folder, 'model', 'dialog', get_params_file_name(self.conf_dict) + '.conf')
save_confs_val(self.conf_dict, self.conf_path)
# set corpus reader
if mode == 'train' :
self.cr = CorpusReaderDialog(dataset_file=dataset_file,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
else :
self.cr = CorpusReaderDialog(dataset_file=None,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
# set model
self.model = BiEncoderAttentionDecoderNetwork(n_words=len(self.cr.get_word_dictionary()),
hidden_status_dim=self.conf_dict['hidden_dim'],
word_embedding_dim=self.conf_dict['word_embedding_dim'],
input_params=param_dict)
# create param sets
if mode == 'test_all' :
self.model_sets, self.model_paths = list(), list()
for parent, dirnames, filenames in os.walk(os.path.join(dataset_folder, 'model', 'dialog')) :
for filename in filenames :
if re.match(algo_name + '_(\w+).model[0-9]+', filename) != None :
self.model_paths.append(filename)
param_dict = load_params_val(os.path.join(dataset_folder, 'model', 'dialog', filename))
model = BiEncoderAttentionDecoderNetwork(n_words=len(self.cr.get_word_dictionary()),
hidden_status_dim=self.conf_dict['hidden_dim'],
word_embedding_dim=self.conf_dict['word_embedding_dim'],
input_params=param_dict)
self.model_sets.append(model)
def train(self):
"""
Train a model.
"""
if self.conf_dict['shuffle'] :
self.cr.shuffle() # The data may be shuffle by some implement of CorpusReader, but NOT ALL.
n_train_set, n_valid_set, n_test_set = self.cr.get_size()
n_batches = (n_train_set - 1) / self.conf_dict['batch_size'] + 1
print 'Compiling model'
train_model = self.model.get_training_function(self.cr, batch_size=self.conf_dict['batch_size'],
batch_repeat=1)
valid_model = self.model.get_validing_function(self.cr)
test_model, pr_model = self.model.get_testing_function(self.cr)
print ('Start to train.')
zae, sae, bae = test_model()
print ('Now testing model. Test Zi Average Error: %s' % (str(zae)))
print ('Now testing model. Test Sentence Average Error: %s' % (str(sae)))
print ('Now testing model. Test Batch Average Error: %s' % (str(bae)))
epoch = 0
n_epochs = 1000000
it = 0
while (epoch < n_epochs):
epoch += 1
for i in xrange(n_batches):
# train model
train_error = train_model(i)[0]
# print 'Step error: %f\r' % train_error,
if math.isnan(train_error):
print 'Train error is NaN in iteration %d' % i
error_model_path = self.param_path + str(epoch) + '.error'
save_params_val(error_model_path, self.model.get_parameters())
print 'model saved in %s , reload and skip the %d batch' % (error_model_path, self.conf_dict['save_freq'])
scope = (self.conf_dict['batch_size'] * i, self.conf_dict['batch_size'] * (i + 1))
for j in range(scope[0], scope[1]):
print '%s\t%s' % (self.cr.train_set[0][j], self.cr.train_set[1][j])
param_dict = load_params_val(self.param_path)
self.model.set_parameters(param_dict)
exit()
# continue
if it % self.conf_dict['save_freq'] == 0:
valid_error = valid_model()[0]
# valid model
print ('@iter: %s\tTraining Error: %s\tValid Error: %s.' %
(it, str(train_error), str(valid_error)))
# Save model parameters
print ('Saving parameters to %s.' % (self.param_path))
save_params_val(self.param_path, self.model.get_parameters())
it = it + 1
# test model
print 'Finished a epoch.'
save_params_val(self.param_path + str(epoch), self.model.get_parameters())
zae, sae, bae = test_model()
print ('Now testing model. Test Zi Average Error: %s' % (str(zae)))
print ('Now testing model. Test Sentence Average Error: %s' % (str(sae)))
print ('Now testing model. Test Batch Average Error: %s' % (str(bae)))
'''
pr_error = pr_model()[0]
test_pr = 0.0
n_samples = config.globalTestPRSamples()
n_data = pr_error.shape[0] / n_samples
trues = [0.0] * (n_samples / 2) + [1.0] * (n_samples / 2)
for i in range(n_data) :
test_pr += pearsonr(trues, list(pr_error[i * n_samples:(i + 1) * n_samples]))[0]
test_pr /= n_data
print ('Now testing model. Test PR: %s' % (str(test_pr)))
'''
print ('\n')
import cPickle
import math
from deep.util.parameter_operation import load_params_val
p = load_params_val('data/human/model/dialog/StyleEncoderDecoderMulti_b498fc540bd33b3d2b33077573e8efe6.model.final')
# def check(l):
# if isinstance(l,list):
# for i in l:
# c = check(i)
# if c:
# return c
# return False
# else:
# return math.isnan(l)
# for k,v in p.items():
# print k, check(v.tolist())
print p
# with open('error', 'wb') as fw:
# cPickle.dump(p, fw)
def __init__(self, dataset_folder, dataset_file, dict_file, stopwords_file,
word_embedding_file, train_rate, valid_rate, test_rate,
algo_name, charset, mode):
"""
Need to set these attributes.
1. conf_dict: configuration of the model.
2. cr: CorpursReader for operate data.
3. model: the network model.
"""
self.conf_dict = {
'algo_name': algo_name,
'batch_size': 256,
'train_valid_test_rate': [train_rate, valid_rate, test_rate],
'split_level': 'zi',
'pre_word_embedding': False,
'word_embedding_dim': 128,
'n_topics': 5000,
'topic_embedding_dim': 256,
'max_sentence_word_num': 150,
'min_sentence_word_num': 1,
'is_BEG': False,
'is_END': True,
'hidden_dim': 512,
'charset': charset,
'shuffle': False,
'save_freq': 100
}
self.param_path = os.path.join(
dataset_folder, 'model', 'dialog',
get_params_file_name(self.conf_dict) + '.model')
print self.param_path
#self.param_path ='ChoEncoderDecoderTopic_5908276eb2ae513520ca72135e5b82d0.model83'
#self.param_path='ChoEncoderDecoderDT_4575b6c5893c10a009e29b6eb2988387.model42'
#self.param_path='ChoEncoderDecoderDT_cc7f5ed5d9e9fe5a90a012f4e017106a.model'
param_dict = load_params_val(self.param_path)
self.conf_path = os.path.join(
dataset_folder, 'model', 'dialog',
get_params_file_name(self.conf_dict) + '.conf')
save_confs_val(self.conf_dict, self.conf_path)
# set corpus reader
if mode == 'train':
self.cr = CorpusReaderDialogTopic(
dataset_file=dataset_file,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
else:
self.cr = CorpusReaderDialogTopic(
dataset_file=None,
stopwords_file=stopwords_file,
dict_file=dict_file,
word_embedding_file=None,
train_valid_test_rate=self.conf_dict['train_valid_test_rate'],
charset=self.conf_dict['charset'],
max_sentence_word_num=self.conf_dict['max_sentence_word_num'],
min_sentence_word_num=self.conf_dict['min_sentence_word_num'],
is_BEG_available=self.conf_dict['is_BEG'],
is_END_available=self.conf_dict['is_END'])
# set model
self.model = RnnEncoderDecoderNetwork(
n_words=len(self.cr.get_word_dictionary()),
hidden_status_dim=self.conf_dict['hidden_dim'],
word_embedding_dim=self.conf_dict['word_embedding_dim'],
n_topics=self.conf_dict['n_topics'],
topic_embedding_dim=self.conf_dict['topic_embedding_dim'],
input_params=param_dict)
