AttributeError: 'SSD' object has no attribute 'module'

[lzx1413]

AttributeError: 'SSD' object has no attribute 'module'

[lzx1413]

You can replace net.module.size with net.size. Module will be added if you train the network with multi-gpus.

[Wondersui]

I have solved it,thank you^ ^

[Wondersui]

I have another problem,when I use the train_test.py to train the voc dataset,it will stop in the Epoch3

Epoch:3 || epochiter: 1974/2068|| Totel iter 6110 || L: 1.5484 C: 3.7036||Batch time: 0.4011 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 1984/2068|| Totel iter 6120 || L: 1.4901 C: 3.6711||Batch time: 0.4020 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 1994/2068|| Totel iter 6130 || L: 1.4994 C: 3.7394||Batch time: 0.4352 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2004/2068|| Totel iter 6140 || L: 1.6116 C: 3.2869||Batch time: 0.4009 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2014/2068|| Totel iter 6150 || L: 1.5345 C: 3.8280||Batch time: 0.3992 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2024/2068|| Totel iter 6160 || L: 1.4239 C: 3.6214||Batch time: 0.4073 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2034/2068|| Totel iter 6170 || L: 1.5346 C: 3.4771||Batch time: 0.4425 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2044/2068|| Totel iter 6180 || L: 1.4747 C: 3.6433||Batch time: 0.4029 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2054/2068|| Totel iter 6190 || L: 1.5019 C: 3.6043||Batch time: 0.4037 sec. ||LR: 0.00050000
Epoch:3 || epochiter: 2064/2068|| Totel iter 6200 || L: 1.3754 C: 3.5850||Batch time: 0.3989 sec. ||LR: 0.00050000

like this,it can not go on.what is the problem about this?

[lzx1413]

I haven't met this problem. What's you pytorch version?

[Wondersui]

I used the pytorch 0.3.1 to run the code.

[lzx1413]

@Wondersui
add

import cv2
cv2.setNumThreads(0)  # pytorch issue 1355: possible deadlock in dataloader

may solve the problem