Can't pickle local object 'TrainAugmentation.__init__.<locals>.

[ZLeopard]

w.start()

File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\process.py", line 105, in start
self._popen = self._Popen(self)
File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\context.py", line 223, in _Popen
return _default_context.get_context().Process._Popen(process_obj)
File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\context.py", line 322, in Popen
return Popen(process_obj)
File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\popen_spawn win32.py", line 65, in init
reduction.dump(process_obj, to_child)
File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\reduction.py ", line 60, in dump
ForkingPickler(file, protocol).dump(obj)
AttributeError: Can't pickle local object 'TrainAugmentation.init..< lambda>'
2018-11-20 16:16:08,607 - root - INFO - working_dir is ./events/ Please using co mmand : tensorboard --logdir=.
Traceback (most recent call last):
File "", line 1, in
File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\spawn.py", l ine 105, in spawn_main
exitcode = _main(fd)
File "D:\SoftWare\Anaconda3\envs\deeplearning\lib\multiprocessing\spawn.py", l ine 115, in _main
self = reduction.pickle.load(from_parent)
EOFError: Ran out of input

have you meet this problem related to the "pickle"?? thank you for the code, firstly~~

[ZLeopard]

ooooooh, i know the reason, just because the OS is Windows, In Windows, multiprocessing uses pickle to transfer objects between processes. Socket objects can not be pickled !!! It's ok in linux~

[ygean]

@ZLeopard Hey, bro! Where and what should I change codes if I want to run on windows platform, I'd like to know your solutions or you just run on linux

[ZLeopard]

@zhouyuangan no, i have no idea about using the code in windows, you may search on the StackOverFlow, some like replace the pickle lib using another lib

[OakSuen]

Hi, have you found the answer how to run the code in Windows, without this error?

[FanLu1994]

anybody solved this problem?

[lianxxx]

It looks some problem with multi-processing, how about setting the num_workers = 0.

[drcdr]

I'm using Windows10 and recent pytorch. I was able to get things to run by :
(1) using num_workers = 0; or,
(2) using num_workers = 4, and:
(2a) replacing
lambda img, boxes=None, labels=None: (img / std, boxes, labels),
with a call to
ScaleStd(std),
in the 3 appropriate places in TrainAugmentation in data_preprocessing.py, and adding an appropriate ScaleStd function to transforms.py.

(2b) for safety, I also added:

    if __name__ == '__main__':
        main()

in train_ssd, see https://discuss.pytorch.org/t/if-name-main-for-window10/19377

Using num_workers=0: [~77s/100 steps]
2019-02-26 00:04:29,682 - root - INFO - Epoch: 0, Step: 100, Average Loss: 10.2757, Average Regression Loss 2.7290, Average Classification Loss: 7.5467
2019-02-26 00:05:47,321 - root - INFO - Epoch: 0, Step: 200, Average Loss: 7.1005, Average Regression Loss 2.3017, Average Classification Loss: 4.7987

Using num_workers=4: [~25s/100 steps]
2019-02-26 00:09:41,438 - root - INFO - Epoch: 0, Step: 100, Average Loss: 10.2432, Average Regression Loss 2.7734, Average Classification Loss: 7.4699
2019-02-26 00:10:05,654 - root - INFO - Epoch: 0, Step: 200, Average Loss: 6.9949, Average Regression Loss 2.2865, Average Classification Loss: 4.7084

So, ~3x speedup (in both cases, I'm reading files from an SSD drive).

I made a few other changes, but I think they are inconsequential. Let me know if this doesn't work for you, and I can check for other possible changes.

[drcdr]

@ZLeopard @zhouyuangan @OakSuen @CodeForWuyu @lianxxx updating you, as this issue is marked closed.

[littlerain2310]

can u find other solution ? It doesn't seem to work on my laptop