There are two loss functions: L
and M
, the total loss is T=L+M
. Is there any difference between computing
L.backward()
M.backward()
optimizer.step()
and
T=L+M
T.backward()
optimizer.step()
There are two loss functions: L
and M
, the total loss is T=L+M
. Is there any difference between computing
L.backward()
M.backward()
optimizer.step()
and
T=L+M
T.backward()
optimizer.step()
The main difference will be that the second one will be faster to get the same result