I want to transfer the kernel parameters in each iteration into the regularization term part. I use the following code.
model=MNISTnet()
print(model.state_dict())
for i, j in model.named_parameters():
print(i)
print(j[0][0][0])
break
The result is shown above. I found that the named_parameters() function get different kernel parameters. The last 2 weights are results of named_parameters() function. I don’t know why they get parameters by re-initialize the kernel rather than the training parameters in each iteration. Please give me some suggestions about this problem, thank you