break
if int(''.join(curentday.split('-'))) == int(''.join("2019-10-27".split('-'))):
curentday = datetime.datetime.strptime(curentday,"%Y-%m-%d") + datetime.timedelta(days=1)
curentday = curentday.strftime("%Y-%m-%d")
continue
cnt = 0
test_data_dict ={}
try:
applog = os.path.join(PATH,logfile.format(curentday))
viewclickfile = handle_log.log(applog, curentday)
#从两个TXT 提取 csv
test_base_path = handle_train.generate_train_base(curentday,viewclickfile)
#train_base = os.path.join(PATH,"train_base/train_base--2019-09-01.csv")
# 结合 mysql数据中 储存的数据，出 训练数据
# maybe we can read ourselves to avoid OOM error!
test_base = pd.read_csv(test_base_path)
#训练 有drop
test_base.drop(labels=["time", "rank", "tag", "cnt"], axis=1, inplace=True)
# print(len(train_base))
# """to test"""
# train_base = train_base[:1]
#def proc_test(train_base):
assert len(test_base) > 0, "len(train_base) must >0!"
if int(''.join(curentday.split('-'))) >= int(''.join("2019-11-06".split('-'))):
logging.info("****Day_nums-1 == index {}****".format(curentday ))
break
if int(''.join(curentday.split('-'))) == int(''.join("2019-10-27".split('-'))):
curentday = datetime.datetime.strptime(curentday,"%Y-%m-%d") + datetime.timedelta(days=1)
curentday = curentday.strftime("%Y-%m-%d")
continue
try:
iiter =0
#从flaskapp log日志中，.1 .2 两个文件，提取出 两个 txt
applog = os.path.join(PATH,logfile.format(curentday))
viewclickfile = handle_log.log(applog, curentday)
#从两个TXT 提取 csv
train_base = handle_train.generate_train_base(curentday,viewclickfile)
#train_base = os.path.join(PATH,"train_base/train_base--2019-09-01.csv")
# 结合 mysql数据中 储存的数据，出 训练数据
train_data = feature.get_data_with_pandas(train_base, batch_size)
lr = 0.001
loss_sum = 0.0
accuracy_sum = 0.0
break_cnt = 1
flag = True
#一天就93个数据吗，每次取一个batch
for i in range(epoches):
for data_pre in train_data:
serve_iter = 1
save_iter = 1
decay_iter = 1
print("*" * 40)
try:
import handle_log
begin = time.time()
handle_log.log(file, day)
print(day, "handle log costs: ", time.time() - begin)
import handle_train
begin = time.time()
train_base = handle_train.generate_train_base(day)
print(day, "handle train_base costs: ", time.time() - begin)
import feature
begin = time.time()
train_data = feature.get_data_with_pandas(train_base, 128)
print(day, "handle feature costs: ", time.time() - begin)
import model
begin = time.time()
mol = model.SimpleModel()
with tf.Session(graph=mol.graph) as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
