How do I change optimizer from adam to sgd, it seems they have different parameters; Because Adam decreases fast and SGD can lead to better result. So I want to train with Adam at first, then use SGD for fine tuning
you can create a new optimizer after a few epochs of training.
Does the model.zero_grad() and optimizer.zero_grad() have same result?
yes, it will. the new optimizer will also be given model.parameters()
so it’s the same result.
1 Like