pytorch 优化器 TypeError 'collections.OrderedDict' 对象不可调用

Question

I used python3.8, pytorch suddenly reported an error optimizer TypeError, but the program was still running two weeks ago.

net = Net(num_classes=7)
net.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(net.parameters(), lr=1e-4)

It seems to be net.parameters() cannot return the correct result?

Traceback (most recent call last):
   File "C:/Users/usr/Desktop/jaffeAttention/jaffe.py", line 190, in <module>
       main()   
   File "C:/Users/usr/Desktop/jaffeAttention/jaffe.py", line 87, in main
      optimizer = optim.Adam(net.parameters(), lr=1e-4)
TypeError: 'collections.OrderedDict' object is not callable
 

enter image description here

Answer 1

You get the error because net.parameters is an OrderedDict , and you then try to call it with net.parameters() , but that is not possible. So you get an error.

To find this out, could try the commands

net = Net(num_classes=7)
net.to(device)
criterion = nn.CrossEntropyLoss()
params= net.parameters
print(type(params)) # verify this is indeed an OrderedDict
optimizer = optim.Adam(params, lr=1e-4) # notice there is no () when passing the params to Adam

You can read about OrderedDict in the python documentation. https://docs.python.org/3/library/collections.html#collections.OrderedDict