An error is reported when the trained model is tested

[SeptemberN]

I saved the model using torch.save() in slover_encoder.py, but I get an error when loading it in conversion.ipynb. None of the model parameters have been changed, why? torch.save(self.G.state_dict(), 'g.ckpt') The error is as follows:

RuntimeError: Error(s) in loading state_dict for Generator:
    size mismatch for encoder.lstm.weight_ih_l0: copying a param with shape torch.Size([64, 512]) from checkpoint, the shape in current model is torch.Size([128, 512]).
    size mismatch for encoder.lstm.weight_hh_l0: copying a param with shape torch.Size([64, 16]) from checkpoint, the shape in current model is torch.Size([128, 32]).
    size mismatch for encoder.lstm.bias_ih_l0: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128]).
    size mismatch for encoder.lstm.bias_hh_l0: copying a param with shape torch.Size([64]) from checkpoint, the shape in current model is torch.Size([128]).
    size mismatch for encoder.lstm.weight_ih_l0_reverse: copying a param with shape torch.Size([64, 512]) from checkpoint, the shape in current model is torch.Size([128, 512]).
    size mismatch for encoder.lstm.weight_hh_l0_reverse: copying a param with shape torch.Size([64, 16]) from checkpoint, the shape in current model is torch.Size([128, 32]).
        ...

[SeptemberN]

Sorry, I made a mistake in the parameters in the program myself. It has now been resolved.

[dnjdsxor21]

@SeptemberN Hello, I got same error with you, using code: torch.save({'model': self.G.state_dict(), 'optimizer': self.g_optimizer.state_dict()}, 'autovc.ckpt')

How did you remove 'the size mismatch error'?

[SeptemberN]

Hi！ You can try to check and modify the added parameters of "parser" in main.py. The shape of the trained model should be consistent with the input data. Tips: You can run 100 saves and test if the model is available