self-defination reader
def make_reader(txt_file, is_train = True):
def reader():
with open(txt_file, 'r') as f:
lines = f.readlines()
f.close()
if is_train:
np.random.shuffle(lines)
# do somehting to balance the samples
for line in lines:
yield preprocess(line, is_train) # self-defination pre-processing
return paddle.reader.buffered(reader,size=256)- for-loop style
db_test = make_reader(test_file, is_train=False)
for i, data in enumerate(db_test()):
# do something meaningful- paddle.batch(reader, batch_size)
my_reader = make_reader(test_file, is_train=False)
test_reader = paddle.batch(my_reader, batch_size=bz) # iterator
# case-1
for idx, data in enumerate(test_reader()):
# type(data) = list, len(data) = batch_size
# case-2
test_py_reader = fluid.layers.py_reader(
capacity=256,
shapes=((-1, 3, 224, 224), (-1, 1)),
lod_levels=[0, 0],
dtypes=('float32', 'int64'),
use_double_buffer=True)
test_py_reader.decorate_paddle_reader(test_reader)
test_py_reader.start()
try:
while True:
test_exe.run(fetch_list=[test_loss.name])
except fluid.core.EOFException:
test_py_reader.reset()fluid.io.PyReader
1
本文地址：https://blog.csdn.net/On_theway10/article/details/110290910
