I am trying to follow NLP From Scratch: Classifying Names with a Character-Level RNN in C++.
First, I would like to know why no optimizer is used like Adam()
? Also, why do we do the following for-loop to add the grad manually?
for (auto const & p : model_to_train.parameters()) {
p.data().add_(p.grad().data() , -k_learning_rate_);
}
Since this loops throws me the error: cannot call variable_data() on undefined tensor (variable_data at ..\..\torch\csrc\autograd\variable.cpp:262)