possible deadlock in dataloader

[nankepan]

HI,
When I train a model with num_workers>1, it is possible to stuck on this line:

DANet/test_DAN.py

Line 137 in f0bc57d

for index, data in enumerate(test_dataloader):

Then I debug and find that it stucks on this tow lines:

DANet/libs/dataset/YoutubeVOS.py

Line 156 in f0bc57d

query_frames, query_masks = self.transforms(query_frames, query_masks)

DANet/libs/dataset/YoutubeVOS.py

Line 157 in f0bc57d

support_frames, support_masks = self.transforms(support_frames, support_masks)

When I train a model when num_workers=0, it is normal but very slow.

The problem is similiar with this: pytorch/pytorch#1355. And I can not fix it using methods under this issue.
How can I fix the problem?

[scutpaul]

hi, if you need to train the model, you should use train_DAN.py.
the default setting for training num_workers is 4

DANet/train_DAN.py

Line 46 in f0bc57d

parser.add_argument('--num_workers', type=int, default=4)

DANet/train_DAN.py

Line 82 in f0bc57d

train_loader = DataLoader(traindataset, batch_size=args.batch_size, shuffle=True, num_workers=args.num_workers,

[nankepan]

I did use train_DAN.py and set num_workers=4.Then sometime it is possible to stuck.

[scutpaul]

hi, you can download our conda yaml to create the python env.
FSVOS.yaml.zip