Questions about model validation

[oceanzhf]

RuntimeError: Error(s) in loading state_dict for Linear:
Missing key(s) in state_dict: "weight", "bias".
Unexpected key(s) in state_dict: "model", "optimizer", "total_steps".
When I use my trained epoch.pth for validate.py, it throws an error. How can I resolve this?

[Bytebender3]

Hello,I have the same problem too.Have you solve it?

[SilkyFinish]

Hello,I have the same problem too.Have you solve it?
I have the same problem too.Have you solve it?

[SilkyFinish]

I have already solved this problem.In the original code the valideate.py needs the pth which only contains two keys:"weight" and "bias",which are the parameters of the final linear layer of the model.However ,after our own training, the pth we saved has three keys:"model", "optimizer", "total_steps",and the "model" contains the paramerters of all the layers in the model.Therfore,you can delete all the other things except the last two parameters of "model" in your pth,which are the needed "weight" and "bias".

[luo3300612]

in validate.py

   state_dict = torch.load(opt.ckpt, map_location='cpu')
   if not ('weight' in state_dict.keys() and 'bias' in state_dict.keys()):
        state_dict = {'weight': state_dict['model']['fc.weight'], 'bias': state_dict['model']['fc.bias']}