Lecture 33: Parameters befor and after training

[huijuanmeng]

In this lecture, we stored the parameters of model before and after training:
params_before = []
for name, p in model.named_parameters():
params_before.append(p.detach().cpu().numpy())

params_after = []
for name, p in model.named_parameters():
params_after.append(p.detach().cpu().numpy())

And we see there are 201 values in each list.

My question is: in the code, we import our saved model, and name it "newmodel", but in "params_after" code, we still get parameters from model, why we don't need to change model name?

Thanks,

[lazyprogrammer]

Thanks for your inquiry.

The variable 'newmodel' is a pipeline object. It's not the same object type as the previous model.

To understand how the parameters stored in the 'model' variable could have changed, you should study "object-oriented programming" to give you some idea of what's happening under the hood.

You should also learn more about Python variables and types. That would be my recommendation.