about loss and optimizer

[wangkwok]

Dear @mishazawa,
Firstly great thanks for your tutorial, it is really very good especially for me, a beginner of ML. Thank you very much!
I have one question about Pytorch. I do not understanding that the loss and optimizer seem NOT link to the model, however only the result y_logits, and label y_train can get the gradient, backward propagation and even update the model parameters? it is so funny!