Can you load a pytorch saved model without knowing its parameters?

[thecaptain2000]

Is it possible to load a saved model without having to communicate the (now useless) parameters which have been used for training it?

at the end of Chapter 3, we exercise with saving and loading the TinyVGG model.
IN the model definition, we do the following:
class TynyVGG(nn.Module):
def init(self, input_shape: int, hidden_units: int, output_shape: int):
...
once the model is trained, the input_shape: hidden_units and output_shape should be redundant (or irrelevant) to the user, as to change them, the user should re train the model. The problem is in order to load the model from file, we need to know those parameters first as we need to:

model_2_loaded = TynyVGG(input_shape=1, hidden_units=30, output_shape=len(class_names)).to(device)

to be able to load the model parameters using:

model_2_loaded.load_state_dict(torch.load(f=MODEL_SAVE_PATH))

Is there a way to load a model without having to communicate the user, the parameters the model was built with?

I was thinking of overloading the constructor and provide a "LoadFromFile" method, is there a better solution?

[shisirkha]

Yeah there is another way but its more complicated than this method you should you this method as the sir said