convert failed with pytorch 0.2.0

[Royzon]

convert failed with pytorch 0.2.0

[unanan]

To load models trained under DataParallel Mode & multi-GPUs, you should modify the code like this:

from collections import OrderedDict

state_dict = torch.load('path/to/YourModel.pth')
new_state_dict = OrderedDict()
for k, v in state_dict.items():
    name = k[7:] # remove `module.`
    new_state_dict[name] = v

pytorch_net.load_state_dict(new_state_dict)