I have followed the swa tutorial of pytorch and saved the state dictionary of the AveragedModel as described here: PyTorch 1.6 now includes Stochastic Weight Averaging | PyTorch
Now, in my standalone inference script, I load the backbone (mobilenet in this case) and try to load the weights of the swa model to it. but I get:
Unexpected key(s) in state_dict: "n_averaged".
I tried finding resources to load swa model state dict but no luck
Any pointers to docs or tips on how to do this would be helpful!