AttributeError: 'DataParallelWithCallback' object has no attribute 'netEMA'

[Ha0Tang]

No description provided.

[edgarschnfld]

Hi,

The problem was caused by the last commit in the updateEMA function:
39fe99b
(using model.netEMA instead of model.module.netEMA).

I fixed it now with a new commit.

[Ha0Tang]

Thanks.

BTW, I cannot find the num_workers parameter in the dataloader.

[edgarschnfld]

The num_workers can be used as argument in the following function:

OASIS/dataloaders/dataloaders.py

Line 23 in 9d4470e

dataloader_train = torch.utils.data.DataLoader(dataset_train, batch_size = opt.batch_size, shuffle = True, drop_last=True)

To make it easier for the user, I now added an optional --num_workers flag in case you don't want to use the default value. For example, using 4 workers looks like this:

python train.py --name oasis_ade20k --dataset_mode ade20k --gpu_ids 0,1 \
--dataroot /some/path/ADEChallengeData2016 --batch_size 32 --num_workers 4

This will automatically set the num_workers for the dataloader function shown above. It's now pushed to master.