Unexpected key(s) in state_dict

1

I get an RuntimeError: Error(s) in loading state_dict for DataParallel: I when I tried to load a model as follows;

model = ModelMark1()
model = torch.nn.DataParallel(model)
checkpoint = torch.load(checkpoint_path, map_location=device)

The details of error;

Unexpected key(s) in state_dict: "module.velocity_model.some_layer.gen_fc.linear.weight", "module.velocity_model.some_layer.gen_fc.linear.bias".

Here is the printed checkpoint dictionary;

odict_keys(['module.spectrogram_extractor.stft.conv_real.weight', 'module.spectrogram_extractor.stft.conv_imag.weight', 'module.logmel_extractor.melW', 'module.bn0.weight', 'module.bn0.bias', 'module.bn0.running_mean', 'module.bn0.running_var', 'module.bn0.num_batches_tracked', 'module.velocity_model.conv_block1.conv1.weight', 'module.velocity_model.conv_block1.conv2.weight', 'module.velocity_model.conv_block1.bn1.weight', 'module.velocity_model.conv_block1.bn1.bias', 'module.velocity_model.conv_block1.bn1.running_mean', 'module.velocity_model.conv_block1.bn1.running_var', 'module.velocity_model.conv_block1.bn1.num_batches_tracked', 'module.velocity_model.conv_block1.bn2.weight', 'module.velocity_model.conv_block1.bn2.bias', 'module.velocity_model.conv_block1.bn2.running_mean', 'module.velocity_model.conv_block1.bn2.running_var', 'module.velocity_model.conv_block1.bn2.num_batches_tracked',  'module.velocity_model.fc.weight', 'module.velocity_model.fc.bias', 'module.velocity_model.some_layer.gen_fc.linear.weight', 'module.velocity_model.some_layer.gen_fc.linear.bias'])

I am not too sure what are the problem here since the module.velocity_model. keys are loaded.
Also the model structure, names are not changed in ModelMark1.
Please let me know if you have advice/clue/solution on this issue.
Thank you so much for your help in advance!

2

Check for typos or provide the state dictionary of the new model you are trying to set.

3

There is no typo I think since it is generated by pytorch and I just loaded them.
The model I trained is one which have those unexpected keys and is the same as ModelMark1. I do not known why there is model does not recognize the 2 keys.
Is there any way to check all expected keys by the ModelMark1?

4

I think the problem is that the layer contains these 2 unexpected keys are not initialized in constructor but in a forward method. Does this cause not creating keys when instantiating the ModelMark1?
If so is there any way to add “mock keys” of the 2 keys in the state_dict and run the model for evaluation or testing purposes?