after loss.backward()，the weight's grad is none

[moonfighting]

When I calculate the loss and run loss.backward,
I found all the weights in the model's grad are None ,
the weights' value can't be updated

the sample codes is like:
`
features = model(input_data)

 loss = circle_loss(features, labels)

 loss.backward()

 for name, param in mode.named_parameters():

    print('name:', name,   "param_grad:', param.grad) # here all the param.grad is None

`