Checking if all layers are used

1

Is it possible to check if all the layers are used during backprogration (basically to see if no layer was forgotten or if it’s output is not connected to the final node) ?

2

param.grad is None when init, backward would make it a variable.

for name,param in model.named_params():
    if param.grad is None:
       print(name)
2 Likes
3

Make sense, thank a lot @chenyuntc !