Model parameter changes on every load

1

Hi,

I saved a model of a simple convnet created without using any loops in the class using torch.save(). When I load the same saved model using torch.load() and print parameters param.data iterating through model.parameters(), it prints different values each time while running the code.

I used torch.cuda.manual_seed(1234) while training the model. But I guess it should not affect loading the saved model.

It might be a silly mistake on my part, but I could not figure it out.

2

Edit: There was a mistake in my code that I overwrote the loading with a model = ConvNet() definition. Please close this thread.