如何使生成器线程安全?
How to make a generator thread-safe?
我有一个看起来像这样的生成器:
def data_generator(data_file, index_list,....):
orig_index_list = index_list
while True:
x_list = list()
y_list = list()
if patch_shape:
index_list = create_patch_index_list(orig_index_list, data_file, patch_shape,
patch_overlap, patch_start_offset,pred_specific=pred_specific)
else:
index_list = copy.copy(orig_index_list)
while len(index_list) > 0:
index = index_list.pop()
add_data(x_list, y_list, data_file, index, augment=augment, augment_flip=augment_flip,
augment_distortion_factor=augment_distortion_factor, patch_shape=patch_shape,
skip_blank=skip_blank, permute=permute)
if len(x_list) == batch_size or (len(index_list) == 0 and len(x_list) > 0):
yield convert_data(x_list, y_list, n_labels=n_labels, labels=labels, num_model=num_model,overlap_label=overlap_label)
x_list = list()
y_list = list()
我的数据集大小为 55GB,存储为 .h5 文件 (data.h5)。读取数据时非常慢。一个纪元需要 7000 秒,我在大约 6 个纪元后出现分段错误。
我想如果我设置 multi_processing = False
和 workers > 1
它会加快读取数据的速度:
model.fit(multi_processing = False, workers = 8)
但是当我这样做时,出现以下错误:
RuntimeError: Your generator is NOT thread-safe. Keras requires a thread-safe generator when use_multiprocessing=False, workers > 1.
有没有办法让我的生成器线程安全?或者有没有其他有效的方法来生成这些数据?
我认为我在上面的评论中引用的 LockedIterator
class 是不正确的,应该按照下面示例中的编码:
import threading
class LockedIterator(object):
def __init__(self, it):
self.lock = threading.Lock()
self.it = iter(it)
def __iter__(self): return self
def __next__(self):
with self.lock:
return self.it.__next__()
def gen():
for x in range(10):
yield x
new_gen = LockedIterator(gen())
def worker(g):
for x in g:
print(x, flush=True)
t1 = threading.Thread(target=worker, args=(new_gen,))
t2 = threading.Thread(target=worker, args=(new_gen,))
t1.start()
t2.start()
t1.join()
t2.join()
打印:
0
1
23
4
5
6
7
8
9
如果要保证打印输出每行打印一个值,那么我们还需要向每个线程传递一个threading.Lock
实例并发出print
该锁控制下的语句,因此打印是序列化的。
我有一个看起来像这样的生成器:
def data_generator(data_file, index_list,....):
orig_index_list = index_list
while True:
x_list = list()
y_list = list()
if patch_shape:
index_list = create_patch_index_list(orig_index_list, data_file, patch_shape,
patch_overlap, patch_start_offset,pred_specific=pred_specific)
else:
index_list = copy.copy(orig_index_list)
while len(index_list) > 0:
index = index_list.pop()
add_data(x_list, y_list, data_file, index, augment=augment, augment_flip=augment_flip,
augment_distortion_factor=augment_distortion_factor, patch_shape=patch_shape,
skip_blank=skip_blank, permute=permute)
if len(x_list) == batch_size or (len(index_list) == 0 and len(x_list) > 0):
yield convert_data(x_list, y_list, n_labels=n_labels, labels=labels, num_model=num_model,overlap_label=overlap_label)
x_list = list()
y_list = list()
我的数据集大小为 55GB,存储为 .h5 文件 (data.h5)。读取数据时非常慢。一个纪元需要 7000 秒,我在大约 6 个纪元后出现分段错误。
我想如果我设置 multi_processing = False
和 workers > 1
它会加快读取数据的速度:
model.fit(multi_processing = False, workers = 8)
但是当我这样做时,出现以下错误:
RuntimeError: Your generator is NOT thread-safe. Keras requires a thread-safe generator when use_multiprocessing=False, workers > 1.
有没有办法让我的生成器线程安全?或者有没有其他有效的方法来生成这些数据?
我认为我在上面的评论中引用的 LockedIterator
class 是不正确的,应该按照下面示例中的编码:
import threading
class LockedIterator(object):
def __init__(self, it):
self.lock = threading.Lock()
self.it = iter(it)
def __iter__(self): return self
def __next__(self):
with self.lock:
return self.it.__next__()
def gen():
for x in range(10):
yield x
new_gen = LockedIterator(gen())
def worker(g):
for x in g:
print(x, flush=True)
t1 = threading.Thread(target=worker, args=(new_gen,))
t2 = threading.Thread(target=worker, args=(new_gen,))
t1.start()
t2.start()
t1.join()
t2.join()
打印:
0
1
23
4
5
6
7
8
9
如果要保证打印输出每行打印一个值,那么我们还需要向每个线程传递一个threading.Lock
实例并发出print
该锁控制下的语句,因此打印是序列化的。