About next(model.parameters())

[ChunZhuo]

torch.manual_seed(42)
model_1 = LinearRegressionModel()
model_1, model_1.state_dict()

OUTPUT:
(LinearRegressionModel(),
OrderedDict([('weights', tensor([0.1940], device='cuda:0')),
('bias', tensor([0.1391], device='cuda:0'))]))

next(model_1.parameters()).device
p1 = next(model_1.parameters())
p2 = next(model_1.parameters())
print(f"There are the params {p1}, {p2}")
print(type(model_1.parameters()))

OUTPUT:
There are the params Parameter containing:
tensor([0.1940], device='cuda:0', requires_grad=True), Parameter containing:
tensor([0.1940], device='cuda:0', requires_grad=True)
<class 'generator'>

Why I cannot get bias using p2 = next(model_1.parameters())?
next function can be used for every object class in pytorch?

[adamomeister]

Are you trying to view your model's weights and biases? If so, I think you need to use the state_dict() method.

[ChunZhuo]

Yes, using state_dict() can work. I also do not understand that next() function can iterate over all the parameters, but why I can only get the weights? I think if I use next() method twice, I can get weights and bias also.

…

________________________________ From: Chris Adamo ***@***.***> Sent: Tuesday, May 9, 2023 10:30 PM To: mrdbourke/pytorch-deep-learning ***@***.***> Cc: Zhang, Joe ***@***.***>; Author ***@***.***> Subject: Re: [mrdbourke/pytorch-deep-learning] About next(model.parameters()) (Discussion #439) Are you trying to view your model's weights and biases? If so, I think you need to use the state_dict() method. — Reply to this email directly, view it on GitHub<#439 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A4FGLFDGX5RLH7JHRZGTZP3XFKSNNANCNFSM6AAAAAAXYFBFSU>. You are receiving this because you authored the thread.Message ID: ***@***.***>