I have trained my model on GPU but I was getting this error while training
AttributeError:
‘collections.OrderedDict’ object has no attribute ‘to’ ,
the code comes after this
my_resnet18 =torch.load(’…/resnet18model/best_model.pth’)
my_resnet18.fc =torch.nn.Sequential(nn.Linear(512, 16),nn.Tanh())
my_resnet18 = my_resnet18.to(device)
Is there anything wrong with this?
my pytorch version is 1.0.0
The problem has been solved thanku
1 Like
How did it work out?
can you tell me how to solve it?
Most likely torch.load()
returned a state_dict
, which would create the issue, if you are trying to call to()
on the OrderedDict
.
The proper way of restoring the model is to initialize the model and load the state_dict
afterwards:
model = models.resnet18()
state_dict = torch.load(...)
model.load_state_dict(state_dict)
model.to(device)
CC @dong-mvp
1 Like