01_Pytorch_Workflow ~05:34 y_preds = model_0(X_test)

[ozzyoz14]

y_preds = model_0(X_test)

I can't understand why/how does this work; I must have a gap in my understanding of classes. If forward() is a method defined under the class, and model_0 is just an instance of that class, why is this not y_preds = model_0.forward(X_test). How does simply listing the input tensor cause forward() to act on it?

[mrdbourke]

Yes you're right it is a little confusing.

However, behind the scenes every subclass of torch.nn.Module calls forward() by default when it's called.

In regular Python classes this kind of functionality might go behind a __call__() method, see more here: https://stackoverflow.com/questions/9663562/what-is-the-difference-between-init-and-call

However, for some reason the PyTorch developers chose to override this functionality with forward(), see more here: https://pytorch.org/docs/stable/generated/torch.nn.Module.html#torch.nn.Module.forward