Hello,
While making a very custom NN, I had to take into account in the forward the fact of being in train mode or not.
I discovered that I could use the function model.train(True) , but I don’t know how to access this mode.
I need to get boolean tensor in order to make an if statement in the forward:
def forward(self, x):
is_train = self.mystery_fn()
if (is_train):
…
else:
…
so far I have build my own is_train parameter, in the init and the functions to change it, but I guess using the proper Pytorch approved one would be better.
Best regards
Barth