How optimizer changes weights

1

take look at my code:

optimizer=optim.Adam(net.parameters(),lr=0.001)
 optimizer.step()

how does optimizer changes net parameters? it is passed once and then they don’t have any connection.is there any pointer stuff there?or i missed something while learning pytorch\python.

2

The optimizer will use the references to all passed parameters and update them in the step function.
You could check the id of each parameter in e.g. optimizer.param_groups[0]['params'] and will see that the ids match the parameter ids in the model.

1 Like