Changing dropout after initialization

1 
dropout = nn.Dropout(0.1)

if dropout is not None:
    dropout.p = 0.2

can you do this? can you change the dropout on the fly after you have declared your model class?

1 Like
2

seems like the this should work as Dropout layer uses self.p as parameter for the nn.functional.dropout function
https://pytorch.org/docs/stable/_modules/torch/nn/modules/dropout.html#Dropout

class Dropout(_DropoutNd):
    def forward(self, input):
        return F.dropout(input, self.p, self.training, self.inplace)