Hello,
I wanted to ask whether there is a way to check if code is executed within a with torch.no_grad(): block?
I have the following problem: I am evaluating my model within with torch.no_grad():. However, my model executes loss.backward() in its forward pass since I want to adapt my deployed model on the unlabelled test data. After adapting my model, I want to evaluate it on the same data using my test function but I am getting the following error: “RuntimError: element 0 of tensors does not require grad and does not have a grad_fn”.
I assume this error is thrown because my model is calling loss.backward() within with torch.no_grad():.
Therefore, I need something like this
#Basic structure of my test function
def test(model, dataloader):
with torch.no_grad():
for data, label in dataloader:
out = model(data) # this is calling my forward function of my model
...
#Custom Forward function of my model
def forward_mymodel(self, input):
out = model(x)
if not TODO:InTorchNoGrad:
loss = softmax_entropy(out).mean(0)
loss.backward()
optimizer.step()
optimizer.zero_grad()
I guess, I could just check one of my input tensors and see if it requires grad but I was wondering if there is an elegant solution for this?
I would appreciate if someone could help me.
Thank you! 
