測試用的代碼

下面每一個對應一個jupyter notebook的單元格

import time
from multiprocessing import Process, Queue

def generator():c = 0while True:time.sleep(1.0) # read somethingyield cc += 1

%%timeds = generator()
for i in range(3):item = next(ds)time.sleep(1.0)  # do somethingprint(item)

0
1
2
CPU times: user 3.27 ms, sys: 456 μs, total: 3.73 ms
Wall time: 6.01 s

def kernel(func, q: Queue):ds = func()while True:item = next(ds)q.put(item)

def multi_generator(func):q = Queue()p = Process(target=kernel, args=(func, q))p.start()while True:item = q.get()yield item

在windows下運行這個單元格的話，會一直顯示在運行，

%%timeds = multi_generator(generator)
for i in range(3):item = next(ds)time.sleep(1.0)print(item)

錯誤

運行上面的代碼，后臺會報如下錯

Traceback (most recent call last):File "<string>", line 1, in <module>File "D:\Anaconda\envs\AIE31\lib\multiprocessing\spawn.py", line 105, in spawn_mainexitcode = _main(fd)File "D:\Anaconda\envs\AIE31\lib\multiprocessing\spawn.py", line 115, in _mainself = reduction.pickle.load(from_parent)
AttributeError: Can't get attribute 'kernel' on <module '__main__' (built-in)>

解決方法

1、些處解決方案，是利用Windows 10 下的WSL的Linux系統解決的。
怎樣配置WSL，請參考鏈接https://blog.csdn.net/jasneik/article/details/123782018

直接打開Ubuntu 控制臺，運行jupyter notebook，一般不會像Windows下的會自動跳轉，可以把jupyter notebook含token的鏈接拷貝到瀏覽器打開。類似如下的鏈接

然后把上面的代碼拷，再運行，就OK了。

%%timeds = multi_generator(generator)
for i in range(3):item = next(ds)time.sleep(1.0)print(item)

0
1
2
CPU times: user 9.26 ms, sys: 1.38 ms, total: 10.6 ms
Wall time: 4.06 s

2、可以參考此方面，是把方法寫到臨時文件（我沒有試過）
https://blog.csdn.net/e274794140/article/details/87286190