Clarification on loading weights after early stopping

1

I have written a class/method for early stopping, but I wanted some clarification on how to load weights to after early stopping. So, we have:

model = Net()
optim=Adam(model.params()) # rough syntax
lrs= <some_lrs>

for epoch in epochs:
    # do something and check for validation loss, assuming this has decreased, do:
    torch.save(model.state_dict(), "es_ckpt.pt")
   # check for early_stopping base don iterations of no_change, assuming no_change, load best weights for model:
   stop()

# load best model like so:
model.load_state_dict(torch.load('es_ckpt.pt'))

is this it? Do I need to create a fresh instance of the model or do I simply load the weights to the existing instance like above?

I also presume I do not need optimiser or lrs paramters in this es_ckpt.pt? since I have them already.

I just wnat to make sure there are no gotchas here, since it sounds too simple :slight_smile: