[Sovled] How to use self parameter out of the model?

class TNET(nn.Module):
    def __init__(self):
        self.weights = nn.Parameter(torch.rand(7), requires_grad=True)
    def forward(self, data):
        return rslts

model = TNET() 
model = torch.nn.DataParallel(model).cuda()
#model  = torch.nn.parallel.DistributedDataParallel(model.cuda(), device_ids=[0,1])

for i, (imgs, lbls) in enumerate(train_loader):
    a = model.weights[0]

File “/opt/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py”, line 518, in __ getattr __ type(self).__ name __, name))
AttributeError: ‘DataParallel’ object has no attribute ‘weights’

If I only use model = TNET(), everything works well. But DistributedDataParallel and DataParallel do not have no attribute ‘weights’.

How could I use the self.weights out of the model with DistributedDataParallel?

Try to access model.module.weights[0].

It works.

Thanks for your help.