def changeWord2index(path_word2vec=None,
path_word_dic=None,
path_save=None):
file_word_dic = open(path_word_dic, 'r', encoding='utf-8')
file_word2vec = open(path_word2vec, 'r', encoding='utf-8')
not_found_word2vec = ' '.join(['0.0'] * 300)
word2vec_list = []
line = file_word_dic.readline()
word2index_dic = {}
word2vec_dic = {}
while line:
cur_word = line.split()[0]
cur_index = line.strip('\n\r').split()[1]
word2index_dic[cur_word] = cur_index
line = file_word_dic.readline()
line = file_word2vec.readline()
while line:
#print(line)
line = line.strip('\n\r').split()
cur_key = str(line[0])
cur_value = ' '.join(line[1:])
word2vec_dic[cur_key] = cur_value
line = file_word2vec.readline()
for cur_key in word2index_dic:
try:
cur_value = word2vec_dic[cur_key]
except:
cur_value = not_found_word2vec
print(cur_key, word2index_dic[cur_key])
cur_string = str(word2index_dic[cur_key]) + ' ' + cur_value
word2vec_list.append(cur_string)
DataUtil.save_vector(path_save, word2vec_list, 'w')
def generate_samples(config):
sess_config = tf.compat.v1.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
default_graph = tf.Graph()
with default_graph.as_default():
sess = tf.compat.v1.Session(config=sess_config, graph=default_graph)
logger = logging.getLogger('')
du = DataUtil(config=config)
du.load_vocab()
generator = Model(config=config, graph=default_graph, sess=sess)
generator.build_train_model()
generator.build_generate(max_len=config.train.max_length,
generate_devices=config.train.devices,
optimizer=config.train.optimizer)
generator.init_and_restore(config.train.modelFile)
infile=config.train.src_path
generate_batch= config.train.batch_size
outfile=config.train.out_path
print("begin generate the data and save the negative in %s" % outfile)
generator.generate_and_save(du, infile, generate_batch, outfile)
print("generate the data done!")
def generate_samples(config):
sess_config = tf.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
default_graph = tf.Graph()
with default_graph.as_default():
sess = tf.Session(config=sess_config, graph=default_graph)
logger = logging.getLogger('')
du = DataUtil(config=config)
du.load_vocab(src_vocab=config.src_vocab,
dst_vocab=config.dst_vocab,
src_vocab_size=config.src_vocab_size_a,
dst_vocab_size=config.dst_vocab_size_b)
generator = Model(config=config, graph=default_graph, sess=sess)
generator.build_variational_train_model()
generator.init_and_restore(config.train.modelFile)
print("begin generate the data and save the negative")
generator.generate_and_save(du,
config.train.src_path,
config.train.batch_size,
config.train.t_domain_generated_data,
direction='ab')
generator.generate_and_save(du,
config.train.dst_path,
config.train.batch_size,
config.train.s_domain_generated_data,
direction='ba')
print("generate the data done!")
def save_all_qid2question():
# 读取配置文件
cf = ConfigParser.ConfigParser()
cf.read("../conf/python.conf")
# 加载train.csv文件
train_data = pd.read_csv('%s/train.csv' % cf.get('DEFAULT', 'origin_pt')).fillna(value="") # [:100]
# 加载test.csv文件
test_data = pd.read_csv('%s/test_with_qid.csv' % cf.get('DEFAULT', 'devel_pt')).fillna(value="") # [:100]
# 存储索引文件
qid2question_qid_fp = '%s/qid2question.all.qid' % cf.get('DEFAULT', 'devel_pt')
qid2question_question_fp = '%s/qid2question.all.question' % cf.get('DEFAULT', 'devel_pt')
# 获取qid2question
all_qid2question = BTM.get_all_qid2question(train_data, test_data)
all_qid = []
all_question = []
for qid in all_qid2question:
all_qid.append(qid)
all_question.append(all_qid2question[qid])
# 存储索引
DataUtil.save_vector(qid2question_qid_fp, all_qid, 'w')
DataUtil.save_vector(qid2question_question_fp, all_question, 'w')
def generate_samples(config):
sess_config = tf.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
default_graph = tf.Graph()
with default_graph.as_default():
sess = tf.Session(config=sess_config, graph=default_graph)
logger = logging.getLogger('')
du = DataUtil(config=config)
du.load_vocab()
generator = Model(config=config, graph=default_graph, sess=sess)
generator.build_train_model()
generator.build_generate(max_len=config.train.max_length,
generate_devices=config.train.devices,
optimizer=config.train.optimizer)
generator.init_and_restore(config.train.modelFile)
infile = config.train.src_path
outfile = config.train.out_path
refile = [config.train.dst_path]
generate_batch = config.train.batch_size
print("begin generate the data and save the negative in %s" % outfile)
generator.generate_and_save(du, infile, generate_batch, outfile)
print("generate the data done!")
SARI_results, BLEU_results = process_evaluation_file_multi(
infile, outfile, refile)
logging.info("SARI: {}, BLEU: {}".format(SARI_results, BLEU_results))
'''
def rescale(online_preds_fp):
online_preds = DataUtil.load_vector(online_preds_fp, 'float')
print(PostProcessor.getResultMean(online_preds))
for index in range(len(online_preds)):
score = online_preds[index]
score = PostProcessor.adj(score, te=0.35, tr=0.25)
online_preds[index] = score
print(PostProcessor.getResultMean(online_preds))
DataUtil.save_vector(online_preds_fp + '.rescale', online_preds, 'w')
def saveSmallWord2Vec(path_save, small_word2vec_dict):
word2vec_list = []
count = 0
for key in small_word2vec_dict:
count += 1
if count < 20:
print(key)
cur_string = str(key) + ' ' + str(small_word2vec_dict[key])
word2vec_list.append(cur_string)
DataUtil.save_vector(path_save, word2vec_list, 'w')
def saveNameAndCalc(self, save_filename, mul_feature_name,
feature_calcpearson):
feature_analysis_pt = self.config.get(
'DIRECTORY', 'feature_analysis_pt') + save_filename
merge_string = []
for cur_feature_name, cur_feature_calc in zip(mul_feature_name,
feature_calcpearson):
merge_string.append(
str(cur_feature_name) + ':' + str(cur_feature_calc))
DataUtil.save_vector(feature_analysis_pt, merge_string, 'w')
def random_split_index_offline(config):
question_offline_fp = config.get('DIRECTORY',
'source_pt') + '/question_train_set.txt'
question_offline = open(question_offline_fp, 'r').readlines()
[train, valid] = DataUtil.random_split(range(len(question_offline)),
[0.966, 0.034])
train_fp = config.get('DIRECTORY', 'index_pt') + 'train_996.offline.index'
valid_fp = config.get('DIRECTORY', 'index_pt') + 'valid_034.offline.index'
DataUtil.save_vector(train_fp, train, 'w')
DataUtil.save_vector(valid_fp, valid, 'w')
def train(config):
logger = logging.getLogger('')
"""Train a model with a config file."""
du = DataUtil(config=config)
du.load_vocab()
model = Model(config=config)
model.build_train_model()
sess_config = tf.compat.v1.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
with model.graph.as_default():
saver = tf.compat.v1.train.Saver(var_list=tf.compat.v1.global_variables())
summary_writer = tf.compat.v1.summary.FileWriter(config.train.logdir, graph=model.graph)
# saver_partial = tf.train.Saver(var_list=[v for v in tf.trainable_variables() if 'Adam' not in v.name])
with tf.compat.v1.Session(config=sess_config) as sess:
# Initialize all variables.
sess.run(tf.compat.v1.global_variables_initializer())
try:
# saver_partial.restore(sess, tf.train.latest_checkpoint(config.train.logdir))
# print('Restore partial model from %s.' % config.train.logdir)
saver.restore(sess, tf.train.latest_checkpoint(config.train.logdir))
except:
logger.info('Failed to reload model.')
for epoch in range(1, config.train.num_epochs+1):
for batch in du.get_training_batches_with_buckets():
start_time = time.time()
step = sess.run(model.global_step)
# Summary
if step % config.train.summary_freq == 0:
step, lr, gnorm, loss, acc, summary, _ = sess.run(
[model.global_step, model.learning_rate, model.grads_norm,
model.loss, model.acc, model.summary_op, model.train_op],
feed_dict={model.src_pl: batch[0], model.dst_pl: batch[1]})
summary_writer.add_summary(summary, global_step=step)
else:
step, lr, gnorm, loss, acc, _ = sess.run(
[model.global_step, model.learning_rate, model.grads_norm,
model.loss, model.acc, model.train_op],
feed_dict={model.src_pl: batch[0], model.dst_pl: batch[1]})
logger.info(
'epoch: {0}\tstep: {1}\tlr: {2:.6f}\tgnorm: {3:.4f}\tloss: {4:.4f}\tacc: {5:.4f}\ttime: {6:.4f}'.
format(epoch, step, lr, gnorm, loss, acc, time.time() - start_time))
# Save model
if step % config.train.save_freq == 0:
mp = config.train.logdir + '/model_epoch_%d_step_%d' % (epoch, step)
saver.save(sess, mp)
logger.info('Save model in %s.' % mp)
logger.info("Finish training.")
def random_split_dataset(config):
all_fp = config.get('DIRECTORY',
'dataset_pt') + 'title_content_word.all.csv'
all_data = open(all_fp, 'r').readlines()
all_data = [line.strip('\n') for line in all_data]
[train, valid] = DataUtil.random_split(all_data, [0.966, 0.034])
train_fp = config.get('DIRECTORY',
'dataset_pt') + 'title_content_word.train_996.csv'
valid_fp = config.get('DIRECTORY',
'dataset_pt') + 'title_content_word.valid_034.csv'
DataUtil.save_vector(train_fp, train, 'w')
DataUtil.save_vector(valid_fp, valid, 'w')
def init_from_config(self, config):
# self.model = Model(config)
self.model = Transformer(config, config.test.devices)
self.model.build_test_model()
sess_config = tf.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
self.sess = tf.Session(config=sess_config, graph=self.model.graph)
# Restore model.
self.model.saver.restore(
self.sess, tf.train.latest_checkpoint(config.train.logdir))
self.du = DataUtil(config)
def save_all_wordtoken(cf):
# 加载train.csv文件
train_data = pd.read_csv('%s/train.csv' % cf.get('DEFAULT', 'origin_pt')).fillna(value="") # [:100]
# 加载test.csv文件
test_data = pd.read_csv('%s/test_with_qid.csv' % cf.get('DEFAULT', 'devel_pt')).fillna(value="") # [:100]
# 文件存储路径
all_wt_fp = '%s/all.wordtoken' % cf.get('DEFAULT', 'devel_pt')
# 获取all_wordtoken
all_wt = BTM.get_all_wordtoken(train_data, test_data)
# 存储
DataUtil.save_vector(all_wt_fp, all_wt, 'w')
def generate_index_with_swap(self):
"""
Generate the index file of `train_with_swap.csv`
:return: none
"""
train_index_fp = '%s/train_311.train.index' % self.config.get('DEFAULT', 'feature_index_pt')
train_with_swap_index_fp = '%s/train_311.train_with_swap.index' % self.config.get('DEFAULT', 'feature_index_pt')
train_index = DataUtil.load_vector(train_index_fp, False)
train_index = [int(x) for x in train_index]
offset = 404290
train_swap_index = [x + offset for x in train_index]
train_with_swap_index = train_index + train_swap_index
DataUtil.save_vector(train_with_swap_index_fp, train_with_swap_index, 'w')
def exe_train(sess, data, epoch, batch_size, hf, feature_shape,
train, loss, input_video, y,
bidirectional=False, step=False,modality='rgb' ):
np.random.shuffle(data)
total_data = len(data)
num_batch = int(math.ceil(total_data/batch_size))+1
total_loss = 0.0
for batch_idx in xrange(num_batch):
batch_data = data[batch_idx*batch_size:min((batch_idx+1)*batch_size,total_data)]
tic = time.time()
data_v,data_y = DataUtil.getOversamplingBatchVideoFeature(batch_data,hf,(10,feature_shape[1],feature_shape[2],feature_shape[3]),modality=modality)
if bidirectional:
flag = np.random.randint(0,2)
if flag==1:
data_v = data_v[:,::-1]
data_time = time.time()-tic
tic = time.time()
# print('data_v mean:', np.mean(data_v),' std:', np.std(data_v))
_, l = sess.run([train,loss],feed_dict={input_video:data_v, y:data_y})
run_time = time.time()-tic
total_loss += l
print(' batch_idx:%d/%d, loss:%.5f, data_time:%.3f, run_time:%.3f' %(batch_idx+1,num_batch,l,data_time,run_time))
total_loss = total_loss/num_batch
return total_loss
def generate(config, argv):
data_name = argv[0]
LogUtil.log('INFO', 'data_name=%s' % data_name)
# load data set
if 'offline' == data_name:
# load offline valid dataset index
valid_index_off_fp = '%s/%s.offline.index' % (config.get('DIRECTORY', 'index_pt'),
config.get('TITLE_CONTENT_CNN', 'valid_index_offline_fn'))
valid_index_off = DataUtil.load_vector(valid_index_off_fp, 'int')
valid_index_off = [num - 1 for num in valid_index_off]
source_file_path = config.get('DIRECTORY', 'source_pt') + '/question_train_set.txt'
source_data = load_raw_line_from_file(config, source_file_path, valid_index_off)
elif 'online' == data_name:
source_file_path = config.get('DIRECTORY', 'source_pt') + '/question_eval_set.txt'
source_data = open(source_file_path, 'r').readlines()
else:
source_data = None
feature_file_path = '%s/instance_fs_length.%s.smat' % (config.get('DIRECTORY', 'dataset_pt'), data_name)
feature_file = open(feature_file_path, 'w')
feature_file.write('%d %d\n' % (len(source_data), 4))
for line in source_data:
qid, tc, tw, dc, dw = parse_question_set(line)
feature = list()
feature.append(len(tc))
feature.append(len(tw))
feature.append(len(dc))
feature.append(len(dw))
Feature.save_feature(feature, feature_file)
feature_file.close()
def saveLabel(data_set_name):
config_fp = '../conf/featwheel.conf'
config = ConfigParser.ConfigParser()
config.read(config_fp)
data = pd.read_csv('%s/%s' %
(config.get('DIRECTORY', 'csv_spanish_cleaning_pt'),
data_set_name)).fillna(value="")
labels_pt = '%s/%s.label' % (config.get(
'DIRECTORY', 'label_pt'), config.get('FEATURE', 'offline_rawset_name'))
labels = []
for index, row in data.iterrows():
cur_label = str(row['is_duplicate'])
labels.append(cur_label)
DataUtil.save_vector(labels_pt, labels, 'w')
def __init__(self, data_dir, split):
self.split = split
self.data_dir = data_dir
if split == "train":
self.alldata = DU.load_json(
os.path.join(data_dir, "train_data.json"))
elif split == "test":
self.alldata = DU.load_json(
os.path.join(data_dir, "test_data.json"))
else:
print("split shoud be train or test")
return
self.graph_lists = []
self.graph_labels = []
self.n_samples = len(self.alldata)
self._prepare()
def demo():
'''
使用样例代码
'''
# 读取配置文件
cf = ConfigParser.ConfigParser()
cf.read("../conf/python.conf")
# 加载特征文件
features = Feature.load("%s/feature1.demo.smat" %
cf.get('DEFAULT', 'feature_question_pt'))
# 存储特征文件
Feature.save(
features,
"%s/feature2.demo.smat" % cf.get('DEFAULT', 'feature_question_pt'))
# 合并特征
Feature.merge_col(features, features)
# 获取<问题>特征池中的特征名
Feature.get_feature_names_question(cf)
# 加载索引文件
indexs = Feature.load_index("%s/vali.demo.index" %
cf.get('DEFAULT', 'feature_index_pt'))
# 根据索引对特征采样
features = Feature.sample_row(features, indexs)
# 正负样本均衡化
rate = 0.165
train311_train_indexs_fp = '%s/train_311.train.index' % cf.get(
'DEFAULT', 'feature_index_pt')
train311_train_indexs = Feature.load_index(train311_train_indexs_fp)
train_labels_fp = '%s/train.label' % cf.get('DEFAULT',
'feature_label_pt')
train_labels = DataUtil.load_vector(train_labels_fp, True)
balanced_indexs = Feature.balance_index(train311_train_indexs,
train_labels, rate)
def exe_test(sess,
data,
batch_size,
v2i,
i2v,
hf,
feature_shape,
predict_words,
input_video,
input_captions,
y,
capl=16):
caption_output = []
total_data = len(data)
num_batch = int(round(total_data * 1.0 / batch_size)) + 1
for batch_idx in xrange(num_batch):
batch_caption = data[batch_idx *
batch_size:min((batch_idx + 1) *
batch_size, total_data)]
data_v = DataUtil.getBatchStepVideoFeature(batch_caption, hf,
feature_shape)
data_c, data_y = DataUtil.getBatchTestCaption(batch_caption,
v2i,
capl=capl)
[gw] = sess.run([predict_words],
feed_dict={
input_video: data_v,
input_captions: data_c,
y: data_y
})
generated_captions = DataUtil.convertCaptionI2V(batch_caption, gw, i2v)
for idx, sen in enumerate(generated_captions):
print('%s : %s' % (batch_caption[idx].keys()[0], sen))
caption_output.append({
'image_id': batch_caption[idx].keys()[0],
'caption': sen
})
js = {}
js['val_predictions'] = caption_output
return js
def run_for_set():
start_time = LoggingUtil.log_start_time()
data = DataUtil.prepare_data()
cls = classifier.Classifier()
cls.perform_with_cross_validation(data, load_from_pickle=False)
LoggingUtil.log_end_time(start_time)
def generate(config, argv):
# load valid dataset index
valid_index_fp = '%s/%s.offline.index' % (config.get('DIRECTORY', 'index_pt'),
config.get('TITLE_CONTENT_CNN', 'valid_index_offline_fn'))
valid_index = DataUtil.load_vector(valid_index_fp, 'int')
valid_index = [num - 1 for num in valid_index]
# load topic btm vec
topic_btm_vec = load_topic_btm_vec(config)
# offline / online
data_name = argv[0]
dis_func_names = ["cosine",
"cityblock",
"jaccard",
"canberra",
"euclidean",
"minkowski",
"braycurtis"]
btm_dis_feature_fn = ['vote_fs_btm_dis_%s' % dis_func_name for dis_func_name in dis_func_names]
btm_dis_feature_f = [open('%s/%s.%s.csv' % (config.get('DIRECTORY', 'dataset_pt'),
fn,
data_name), 'w') for fn in btm_dis_feature_fn]
if 'offline' == data_name:
btm_tw_cw_features = load_features_from_file(config, 'fs_btm_tw_cw', data_name, valid_index)
LogUtil.log('INFO', 'load_features_from_file, len=%d' % len(btm_tw_cw_features))
for line_id in range(len(btm_tw_cw_features)):
doc_vec = btm_tw_cw_features[line_id]
for dis_id in range(len(dis_func_names)):
vec = [0.] * 1999
for topic_id in range(1999):
topic_vec = topic_btm_vec[topic_id]
if 'minkowski' == dis_func_names[dis_id]:
vec[topic_id] = eval(dis_func_names[dis_id])(doc_vec, topic_vec, 3)
else:
vec[topic_id] = eval(dis_func_names[dis_id])(doc_vec, topic_vec)
btm_dis_feature_f[dis_id].write('%s\n' % ','.join([str(num) for num in vec]))
else:
btm_vec_fp = '%s/fs_btm_tw_cw.%s.csv' % (config.get('DIRECTORY', 'dataset_pt'), data_name)
btm_vec_f = open(btm_vec_fp, 'r')
for line in btm_vec_f:
doc_vec = np.nan_to_num(parse_feature_vec(line))
for dis_id in range(len(dis_func_names)):
vec = [0.] * 1999
for topic_id in range(1999):
topic_vec = topic_btm_vec[topic_id]
if 'minkowski' == dis_func_names[dis_id]:
vec[topic_id] = eval(dis_func_names[dis_id])(doc_vec, topic_vec, 3)
else:
vec[topic_id] = eval(dis_func_names[dis_id])(doc_vec, topic_vec)
btm_dis_feature_f[dis_id].write('%s\n' % ','.join([str(num) for num in vec]))
for f in btm_dis_feature_f:
f.close()
def save_tmp_grid_jsons(tmp_grid_loc, accumulation_data, raw_data_loc):
for i in range(5):
for j in range(10):
DU.mkdir(
os.path.join(tmp_grid_loc, "p%s" % str(i + 1),
"m%s" % str(j + 1)))
grid_data_info = {}
if os.path.exists(os.path.join(tmp_grid_loc,
"config.json")) and accumulation_data:
grid_data_info = DU.load_json(os.path.join(tmp_grid_loc,
"config.json"))
row_p = grid_data_info['row_p']
col_m = grid_data_info['col_m']
dir_count = grid_data_info['json_count']
else:
row_p = [5 + 2 * (i + 1) for i in range(5)]
col_m = [4 + 4 * (i + 1) for i in range(10)]
dir_count = np.zeros([len(row_p), len(col_m)], dtype=int)
grid_data_info['row_p'] = row_p
grid_data_info['col_m'] = col_m
print(dir_count)
print(row_p)
print(col_m)
all_data_json = DU.load_json(raw_data_loc)
print(len(all_data_json))
# 分栏保存数据
for data in all_data_json.values():
petri_net = data["petri_net"]
v_list = data["arr_vlist"]
p_num = get_lowest_idx(len(petri_net), row_p)
m_num = get_lowest_idx(len(v_list), col_m)
dir_count[p_num - 1][m_num - 1] = dir_count[p_num - 1][m_num - 1] + 1
DU.save_data_to_json(
os.path.join(
tmp_grid_loc, "p%s" % str(p_num), "m%s" % str(m_num),
"data%s.json" % str(int(dir_count[p_num - 1][m_num - 1]))),
data)
print(dir_count)
# print(get_lowest_idx(7,row_p))
if isinstance(dir_count, list):
grid_data_info["json_count"] = dir_count
else:
grid_data_info["json_count"] = dir_count.tolist()
DU.save_data_to_json(os.path.join(tmp_grid_loc, "config.json"),
grid_data_info)
def __init__(self, config):
self.config = config
# Load model
self.model = Model(config)
self.model.build_test_model()
self.du = DataUtil(config)
self.du.load_vocab()
# Create session
sess_config = tf.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
self.sess = tf.Session(config=sess_config, graph=self.model.graph)
# Restore model.
with self.model.graph.as_default():
saver = tf.train.Saver(tf.global_variables())
saver.restore(self.sess, tf.train.latest_checkpoint(config.train.logdir))
def exe_train(sess,
data,
batch_size,
v2i,
hf,
feature_shape,
train,
loss,
input_video,
input_captions,
y,
capl=16):
np.random.shuffle(data)
total_data = len(data)
num_batch = int(round(total_data * 1.0 / batch_size))
total_loss = 0.0
for batch_idx in xrange(num_batch):
# if batch_idx < 100:
batch_caption = data[batch_idx *
batch_size:min((batch_idx + 1) *
batch_size, total_data)]
data_v = DataUtil.getBatchStepVideoFeature(batch_caption, hf,
feature_shape)
data_c, data_y = DataUtil.getNewBatchTrainCaption(batch_caption,
v2i,
capl=capl)
_, l = sess.run([train, loss],
feed_dict={
input_video: data_v,
input_captions: data_c,
y: data_y
})
total_loss += l
print(' batch_idx:%d/%d, loss:%.5f' % (batch_idx + 1, num_batch, l))
total_loss = total_loss / num_batch
return total_loss
def generate(config, argv):
data_name = argv[0]
word_idf_fp = '%s/words.idf' % config.get('DIRECTORY', 'devel_pt')
with open(word_idf_fp, 'r') as word_idf_f:
word_idf = json.load(word_idf_f)
LogUtil.log("INFO", "load word_idf done, len(word_idf)=%d" % len(word_idf))
char_idf_fp = '%s/chars.idf' % config.get('DIRECTORY', 'devel_pt')
with open(char_idf_fp, 'r') as char_idf_f:
char_idf = json.load(char_idf_f)
LogUtil.log("INFO", "load char_idf done, len(char_idf)=%d" % len(char_idf))
# load data set
if 'offline' == data_name:
# load offline valid dataset index
valid_index_off_fp = '%s/%s.offline.index' % (
config.get('DIRECTORY', 'index_pt'),
config.get('TITLE_CONTENT_CNN', 'valid_index_offline_fn'))
valid_index_off = DataUtil.load_vector(valid_index_off_fp, 'int')
valid_index_off = [num - 1 for num in valid_index_off]
source_file_path = config.get('DIRECTORY',
'source_pt') + '/question_train_set.txt'
source_data = load_raw_line_from_file(config, source_file_path,
valid_index_off)
features = valid_index_off
elif 'online' == data_name:
source_file_path = config.get('DIRECTORY',
'source_pt') + '/question_eval_set.txt'
source_data = open(source_file_path, 'r').readlines()
features = range(len(source_data))
else:
source_data = None
features = None
id_feature_file_path = '%s/instance_fs_id.%s.smat' % (config.get(
'DIRECTORY', 'dataset_pt'), data_name)
feature_file = open(id_feature_file_path, 'w')
feature_file.write('%d %d\n' % (len(source_data), 1))
for id_num in features:
feature = list()
feature.append(id_num % 100000)
Feature.save_feature(feature, feature_file)
feature_file.close()
def generate_spn(config, write_loc, data_idx):
place_upper_bound = config['place_upper_bound']
marks_lower_limit = config['marks_lower_limit']
marks_upper_limit = config['marks_upper_limit']
prune_flag = config['prune_flag']
add_token = config['add_token']
max_place_num = config['max_place_num']
min_place_num = config['min_place_num']
finish = False
while finish == False:
place_num = np.random.randint(min_place_num, max_place_num + 1)
tran_num = place_num + np.random.randint(-3, 1)
petri_matrix = PeGen.rand_generate_petri(place_num, tran_num)
if prune_flag:
petri_matrix = PeGen.prune_petri(petri_matrix)
if add_token:
petri_matrix = PeGen.add_token(petri_matrix)
results_dict, finish = SPN.filter_spn(petri_matrix, place_upper_bound,
marks_lower_limit,
marks_upper_limit)
DU.save_data_to_json(
os.path.join(write_loc, "data%s.json" % str(data_idx)), results_dict)
def save_all_question2wordtoken(cf):
# 加载train.csv文件
train_data = pd.read_csv('%s/train.csv' % cf.get('DEFAULT', 'origin_pt')).fillna(value="") # [:100]
# 加载test.csv文件
test_data = pd.read_csv('%s/test_with_qid.csv' % cf.get('DEFAULT', 'devel_pt')).fillna(value="") # [:100]
# 文件存储路径
q2wt_q_fp = '%s/q2wt.all.question' % cf.get('DEFAULT', 'devel_pt')
q2wt_wt_fp = '%s/q2wt.all.wordtoken' % cf.get('DEFAULT', 'devel_pt')
# 获取qid2question
all_q2wt = BTM.get_all_question2wordtoken(train_data, test_data)
all_q = []
all_wt = []
for q in all_q2wt:
all_q.append(q)
all_wt.append(all_q2wt[q])
# 存储索引
DataUtil.save_vector(q2wt_q_fp, all_q, 'w')
DataUtil.save_vector(q2wt_wt_fp, all_wt, 'w')
def __init__(self, config):
self.config = config
# Load model
self.model = Model(config)
# self.model.build_test_model()
self.model.build_variational_test_model(mode=config.test.mode)
logging.info('build_test_variational_model done!')
self.du = DataUtil(config)
self.du.load_vocab(src_vocab=config.src_vocab,
dst_vocab=config.dst_vocab,
src_vocab_size=config.src_vocab_size_a,
dst_vocab_size=config.src_vocab_size_b)
# Create session
sess_config = tf.ConfigProto()
sess_config.gpu_options.allow_growth = True
sess_config.allow_soft_placement = True
self.sess = tf.Session(config=sess_config, graph=self.model.graph)
# Restore model.
with self.model.graph.as_default():
saver = tf.train.Saver(tf.global_variables())
saver.restore(self.sess, tf.train.latest_checkpoint(config.train.logdir))
def generate_cv_subset_index(cf, argv):
"""
Generate index used for 5-fold cross validation
:param cf: configuration file
:param argv: parameter list
:return: none
"""
tag = argv[0]
cv_num = 5
cv_rawset_name = 'train_with_swap'
train_data_size = 404290
index_all = []
for i in range(cv_num):
index_all.append([])
for i in range(train_data_size):
index_all[int(random.random() * cv_num)].append(i)
for i in range(cv_num):
LogUtil.log('INFO', 'size(part%d)=%d' % (i, len(index_all[i])))
index_fp = cf.get('DEFAULT', 'feature_index_pt')
for i in range(cv_num):
fold_id = i
# train
fp = '%s/cv_tag%s_n%d_f%d_train.%s.index' % (
index_fp, tag, cv_num, fold_id, cv_rawset_name)
for j in range(cv_num - 2):
part_id = (i + j) % cv_num
DataUtil.save_vector(fp, index_all[part_id], 'a')
for j in range(cv_num - 2):
part_id = (i + j) % cv_num
DataUtil.save_vector(
fp,
[index + train_data_size
for index in index_all[part_id]], 'a')
# valid
fp = '%s/cv_tag%s_n%d_f%d_valid.%s.index' % (
index_fp, tag, cv_num, fold_id, cv_rawset_name)
part_id = (fold_id + cv_num - 2) % cv_num
DataUtil.save_vector(fp, index_all[part_id], 'w')
# test
fp = '%s/cv_tag%s_n%d_f%d_test.%s.index' % (
index_fp, tag, cv_num, fold_id, cv_rawset_name)
part_id = (fold_id + cv_num - 1) % cv_num
DataUtil.save_vector(fp, index_all[part_id], 'w')
def save_train_qid2question():
# 读取配置文件
cf = ConfigParser.ConfigParser()
cf.read("../conf/python.conf")
# 加载train.csv文件
train_data = pd.read_csv('%s/train.csv' % cf.get('DEFAULT', 'origin_pt')).fillna(value="") # [:100]
# 存储文件路径
qid2question_qid_fp = '%s/qid2question.train.qid' % cf.get('DEFAULT', 'devel_pt')
qid2question_question_fp = '%s/qid2question.train.question' % cf.get('DEFAULT', 'devel_pt')
# 获取qid2question
train_qid2question = BTM.get_qid2question(train_data)
train_qid = []
train_question = []
for qid in train_qid2question:
train_qid.append(qid)
train_question.append(train_qid2question[qid])
# 存储索引
DataUtil.save_vector(qid2question_qid_fp, train_qid, 'w')
DataUtil.save_vector(qid2question_question_fp, train_question, 'w')
def run_for_examples():
start_time = LoggingUtil.log_start_time()
data = DataUtil.prepare_data()
cls = MultinomialNB()
vect = CountVectorizer(ngram_range=(1, 2))
train_labels = data['label'].values
train_features = vect.fit_transform(data['email'].values)
cls.fit(train_features, train_labels)
examples = ['Congrats! Boss is proud of your promotion. Keep doing well. Regards.',
'Congrats! You are lucky one to be offered a promotion!',
'Congrats! You are promoted!',
'Congrats! You won one million!']
test_features = vect.transform(examples)
predictions = cls.predict(test_features)
print(predictions)
LoggingUtil.log_end_time(start_time)