I have a pretrained network parameters weight file, and now I need to add some intermediate layers into the network, but the function model.load_state_dict() doesn’t work for this need. Is there any alternative methods to do this job ?
Thank you !
You could load the trained weights into your “standard model”, while leaving the new changed model empty.
Then you could try to iterate over the state dict and assigning the weights to your new model’s layer.